write license

build(deps): bump ureq from 3.0.12 to 3.1.0

.env.sample

@@ -3,4 +3,5 @@ MODEL_PATH=models/tsukuyomi.sbv2
 MODELS_PATH=models
 TOKENIZER_PATH=models/tokenizer.json
 ADDR=localhost:3000
-RUST_LOG=warn
+RUST_LOG=warn
+HOLDER_MAX_LOADED_MODElS=20

.github/FUNDING.yml

@@ -0,0 +1,3 @@
+# These are supported funding model platforms
+
+github: [tuna2134]

.github/dependabot.yml

@@ -0,0 +1,15 @@
+# To get started with Dependabot version updates, you'll need to specify which
+# package ecosystems to update and where the package manifests are located.
+# Please see the documentation for all configuration options:
+# https://docs.github.com/code-security/dependabot/dependabot-version-updates/configuration-options-for-the-dependabot.yml-file
+
+version: 2
+updates:
+  - package-ecosystem: "cargo" # See documentation for possible values
+    directory: "/" # Location of package manifests
+    schedule:
+      interval: "weekly"
+  - package-ecosystem: "npm" # See documentation for possible values
+    directory: "/" # Location of package manifests
+    schedule:
+      interval: "weekly"

.github/pull_request_template.md

@@ -1,8 +1,13 @@
 ## 概要
-(ここに本PRの説明をしてください。)
+<!--
+ここに本PRの説明をしてください。
+-->
 
 ## 関連issue
-(ここに該当するissueの番号を書いてください。)
+<!--
+ここに該当するissueの番号を書いてください。
+#nの前にfixesを置くとプルリクが閉じた時に自動的に該当issueもクローズします、
+-->
 
 ## 確認
 - [ ] 動作確認しましたか？

.github/workflows/CI.yml

@@ -1,132 +0,0 @@
-# This file is autogenerated by maturin v1.7.1
-# To update, run
-#
-#    maturin generate-ci github
-#
-name: CI
-
-on:
-  push:
-    branches:
-      - main
-      - master
-    tags:
-      - '*'
-  pull_request:
-  workflow_dispatch:
-
-permissions:
-  contents: read
-  id-token: write
-
-jobs:
-  linux:
-    runs-on: ${{ matrix.platform.runner }}
-    strategy:
-      matrix:
-        platform:
-          - runner: ubuntu-latest
-            target: x86_64
-          - runner: ubuntu-latest
-            target: aarch64
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
-        with:
-          python-version: 3.x
-      - name: Build wheels
-        uses: PyO3/maturin-action@v1
-        with:
-          target: ${{ matrix.platform.target }}
-          args: --release --out dist --find-interpreter
-          sccache: 'true'
-          manylinux: auto
-          working-directory: sbv2_bindings
-      - name: Upload wheels
-        uses: actions/upload-artifact@v4
-        with:
-          name: wheels-linux-${{ matrix.platform.target }}
-          path: dist
-
-  windows:
-    runs-on: ${{ matrix.platform.runner }}
-    strategy:
-      matrix:
-        platform:
-          - runner: windows-latest
-            target: x64
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
-        with:
-          python-version: 3.x
-          architecture: ${{ matrix.platform.target }}
-      - name: Build wheels
-        uses: PyO3/maturin-action@v1
-        with:
-          target: ${{ matrix.platform.target }}
-          args: --release --out dist --find-interpreter
-          sccache: 'true'
-          working-directory: sbv2_bindings
-      - name: Upload wheels
-        uses: actions/upload-artifact@v4
-        with:
-          name: wheels-windows-${{ matrix.platform.target }}
-          path: dist
-
-  macos:
-    runs-on: ${{ matrix.platform.runner }}
-    strategy:
-      matrix:
-        platform:
-          - runner: macos-12
-            target: x86_64
-          - runner: macos-14
-            target: aarch64
-    steps:
-      - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
-        with:
-          python-version: 3.x
-      - name: Build wheels
-        uses: PyO3/maturin-action@v1
-        with:
-          target: ${{ matrix.platform.target }}
-          args: --release --out dist --find-interpreter
-          sccache: 'true'
-          working-directory: sbv2_bindings
-      - name: Upload wheels
-        uses: actions/upload-artifact@v4
-        with:
-          name: wheels-macos-${{ matrix.platform.target }}
-          path: dist
-
-  sdist:
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - name: Build sdist
-        uses: PyO3/maturin-action@v1
-        with:
-          command: sdist
-          args: --out dist
-          working-directory: sbv2_bindings
-      - name: Upload sdist
-        uses: actions/upload-artifact@v4
-        with:
-          name: wheels-sdist
-          path: dist
-
-  release:
-    name: Release
-    runs-on: ubuntu-latest
-    if: "startsWith(github.ref, 'refs/tags/')"
-    needs: [linux, windows, macos, sdist]
-    environment: release
-    steps:
-      - uses: actions/download-artifact@v4
-      - name: Publish to PyPI
-        uses: PyO3/maturin-action@v1
-        with:
-          command: upload
-          args: --non-interactive --skip-existing wheels-*/*

.github/workflows/build.Dockerfile

@@ -0,0 +1,4 @@
+FROM ubuntu:latest
+RUN apt update && apt install openssl libssl-dev curl pkg-config software-properties-common -y && add-apt-repository ppa:deadsnakes/ppa && apt update && apt install python3.7 python3.8 python3.9 python3.10 python3.11 python3.12 python3.13 python3-pip python3 -y
+ENV PIP_BREAK_SYSTEM_PACKAGES=1
+RUN mkdir -p /root/.cache/sbv2 && curl https://huggingface.co/neody/sbv2-api-assets/resolve/main/dic/all.bin -o /root/.cache/sbv2/all.bin -L

.github/workflows/build.yml

@@ -1,40 +1,220 @@
-name: Push to github container register
+name: Build
 
 on:
-  release:
-    types: [created]
+  push:
+    branches:
+      - main
+    tags:
+      - '*'
   workflow_dispatch:
 
+permissions:
+  contents: write
+  id-token: write
+  packages: write
+
 jobs:
-  push-docker:
-    runs-on: ubuntu-latest
-    permissions:
-      contents: read
-      packages: write
+  python-linux:
+    runs-on: ${{ matrix.platform.runner }}
     strategy:
       matrix:
-        tag: [cpu, cuda]
         platform:
-        - linux/amd64
-        - linux/arm64
+          - runner: ubuntu-latest
+            target: x86_64
+          - runner: ubuntu-24.04-arm
+            target: aarch64
     steps:
       - uses: actions/checkout@v4
-      - name: Set up QEMU
-        uses: docker/setup-qemu-action@v3
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-      - name: Login to GitHub Container Registry
+      - uses: actions/setup-python@v5
+        with:
+          python-version: 3.x
+      - run: docker build . -f .github/workflows/build.Dockerfile --tag ci
+      - name: Build wheels
+        uses: PyO3/maturin-action@v1
+        with:
+          target: ${{ matrix.platform.target }}
+          args: --release --out dist --find-interpreter
+          sccache: 'true'
+          manylinux: auto
+          container: ci
+          working-directory: ./crates/sbv2_bindings
+      - name: Upload wheels
+        uses: actions/upload-artifact@v4
+        with:
+          name: wheels-linux-${{ matrix.platform.target }}
+          path: ./crates/sbv2_bindings/dist
+
+  python-windows:
+    runs-on: ${{ matrix.platform.runner }}
+    strategy:
+      matrix:
+        platform:
+          - runner: windows-latest
+            target: x64
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: 3.x
+          architecture: ${{ matrix.platform.target }}
+      - name: Build wheels
+        uses: PyO3/maturin-action@v1
+        with:
+          target: ${{ matrix.platform.target }}
+          args: --release --out dist --find-interpreter
+          sccache: 'true'
+          working-directory: ./crates/sbv2_bindings
+      - name: Upload wheels
+        uses: actions/upload-artifact@v4
+        with:
+          name: wheels-windows-${{ matrix.platform.target }}
+          path: ./crates/sbv2_bindings/dist
+
+  python-macos:
+    runs-on: ${{ matrix.platform.runner }}
+    strategy:
+      matrix:
+        platform:
+          - runner: macos-14
+            target: aarch64
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: 3.x
+      - name: Build wheels
+        uses: PyO3/maturin-action@v1
+        with:
+          target: ${{ matrix.platform.target }}
+          args: --release --out dist --find-interpreter
+          sccache: 'true'
+          working-directory: ./crates/sbv2_bindings
+      - name: Upload wheels
+        uses: actions/upload-artifact@v4
+        with:
+          name: wheels-macos-${{ matrix.platform.target }}
+          path: ./crates/sbv2_bindings/dist
+
+  python-sdist:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - name: Build sdist
+        uses: PyO3/maturin-action@v1
+        with:
+          command: sdist
+          args: --out dist
+          working-directory: ./crates/sbv2_bindings
+      - name: Upload sdist
+        uses: actions/upload-artifact@v4
+        with:
+          name: wheels-sdist
+          path: ./crates/sbv2_bindings/dist
+
+  python-wheel:
+    name: Wheel Upload
+    runs-on: ubuntu-latest
+    needs: [python-linux, python-windows, python-macos, python-sdist]
+    env:
+      GH_TOKEN: ${{ github.token }}
+    steps:
+      - uses: actions/checkout@v4
+      - run: gh run download ${{ github.run_id }} -p wheels-*
+      - name: release
+        run: |
+          gh release create commit-${GITHUB_SHA:0:8} --prerelease wheels-*/*
+
+  python-release:
+    name: Release
+    runs-on: ubuntu-latest
+    if: "startsWith(github.ref, 'refs/tags/')"
+    needs: [python-linux, python-windows, python-macos, python-sdist]
+    environment: release
+    env:
+      GH_TOKEN: ${{ github.token }}
+    steps:
+      - uses: actions/checkout@v4
+      - run: gh run download ${{ github.run_id }} -p wheels-*
+      - name: Publish to PyPI
+        uses: PyO3/maturin-action@v1
+        with:
+          command: upload
+          args: --non-interactive --skip-existing wheels-*/*
+
+  docker:
+    runs-on: ${{ matrix.machine.runner }}
+    strategy:
+      fail-fast: false
+      matrix:
+        machine:
+          - platform: amd64
+            runner: ubuntu-latest
+          - platform: arm64
+            runner: ubuntu-24.04-arm
+        tag: [cpu, cuda]
+    steps:
+      - name: Prepare
+        run: |
+          platform=${{ matrix.machine.platform }}
+          echo "PLATFORM_PAIR=${platform//\//-}" >> $GITHUB_ENV
+
+      - name: Docker meta
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: |
+            ghcr.io/${{ github.repository }}
+
+      - name: Login to GHCR
         uses: docker/login-action@v3
         with:
           registry: ghcr.io
-          username: ${{ github.actor }}
+          username: ${{ github.repository_owner }}
           password: ${{ secrets.GITHUB_TOKEN }}
-      - name: Build and push image
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@v3
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Build and push by digest
+        id: build
         uses: docker/build-push-action@v6
         with:
-          context: .
+          labels: ${{ steps.meta.outputs.labels }}
+          file: ./scripts/docker/${{ matrix.tag }}.Dockerfile
           push: true
           tags: |
-            ghcr.io/${{ github.repository }}:${{ matrix.tag }}
-          file: docker/${{ matrix.tag }}.Dockerfile
-          platforms: ${{ matrix.platform }}
+            ghcr.io/${{ github.repository }}:latest-${{ matrix.tag }}-${{ matrix.machine.platform }}
+
+  docker-merge:
+    runs-on: ubuntu-latest
+    needs:
+      - docker
+    steps:
+      - name: Download digests
+        uses: actions/download-artifact@v4
+        with:
+          path: ${{ runner.temp }}/digests
+          pattern: digests-*
+          merge-multiple: true
+
+      - name: Login to GHCR
+        uses: docker/login-action@v3
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v3
+
+      - name: Merge
+        run: |
+          docker buildx imagetools create -t ghcr.io/${{ github.repository }}:cuda \
+              ghcr.io/${{ github.repository }}:latest-cuda-amd64 \
+              ghcr.io/${{ github.repository }}:latest-cuda-arm64
+          docker buildx imagetools create -t ghcr.io/${{ github.repository }}:cpu \
+              ghcr.io/${{ github.repository }}:latest-cpu-amd64 \
+              ghcr.io/${{ github.repository }}:latest-cpu-arm64

.github/workflows/lint.yml

@@ -0,0 +1,26 @@
+name: Lint
+
+on:
+  pull_request:
+
+jobs:
+  check:
+    runs-on: ubuntu-latest
+    strategy:
+      fail-fast: false
+      matrix:
+        components:
+          - rustfmt
+          - clippy
+    steps:
+      - name: Setup
+        uses: actions/checkout@v4
+      - uses: actions-rust-lang/setup-rust-toolchain@v1
+        with:
+          components: ${{ matrix.components }}
+      - name: Format
+        if: ${{ matrix.components == 'rustfmt' }}
+        run: cargo fmt --all -- --check
+      - name: Lint
+        if: ${{ matrix.components == 'clippy' }}
+        run: cargo clippy --all-targets --all-features -- -D warnings

.gitignore

@@ -1,7 +1,10 @@
-target
+target/
 models/
 !models/.gitkeep
 venv/
 .env
-output.wav
-node_modules
+*.wav
+node_modules/
+dist/
+*.csv
+*.bin

Cargo.toml

@@ -1,14 +1,25 @@
 [workspace]
-resolver = "2"
-members = ["sbv2_api", "sbv2_core", "sbv2_bindings"]
+resolver = "3"
+members = ["./crates/sbv2_api", "./crates/sbv2_core", "./crates/sbv2_bindings", "./crates/sbv2_wasm"]
+
+[workspace.package]
+version = "0.2.0-alpha6"
+edition = "2021"
+description = "Style-Bert-VITSの推論ライブラリ"
+license = "MIT"
+readme = "./README.md"
+repository = "https://github.com/neodyland/sbv2-api"
+documentation = "https://docs.rs/sbv2_core"
 
 [workspace.dependencies]
-anyhow = "1.0.86"
+anyhow = "1.0.99"
 dotenvy = "0.15.7"
-env_logger = "0.11.5"
+env_logger = "0.11.6"
 ndarray = "0.16.1"
+once_cell = "1.20.3"
 
 [profile.release]
+strip = true
+opt-level = "z"
 lto = true
-debug = false
-strip = true
+codegen-units = 1

LICENSE

@@ -1,6 +1,7 @@
 MIT License
 
 Copyright (c) 2024 tuna2134
+Copyright (c) 2025- neodyland
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -1,8 +1,23 @@
 # SBV2-API
 
+> [!CAUTION]
+> 本バージョンはアルファ版です。
+> 
+> 安定版を利用したい場合は[こちら](https://github.com/neodyland/sbv2-api/tree/v0.1.x)をご覧ください。
+
+> [!CAUTION]
+> オプションの辞書はLGPLです。
+> 
+> オプションの辞書を使用する場合、バイナリの内部の辞書部分について、LGPLが適用されます。
+
+> [!NOTE]
+> このレポジトリはメンテナンスの都合上、[tuna2134](https:://github.com/tuna2134)氏の所属する[Neodyland](https://neody.land/)へとリポジトリ所在地を移動しました。
+> 
+> 引き続きtuna2134氏がメインメンテナとして管理しています。
+
 ## プログラミングに詳しくない方向け
 
-[こちら](https://github.com/tuna2134/sbv2-gui?tab=readme-ov-file)を参照してください。
+[こちら](https://github.com/tuna2134/sbv2-gui)を参照してください。
 
 コマンドやpythonの知識なしで簡単に使えるバージョンです。(できることはほぼ同じ)
 
@@ -14,7 +29,7 @@ JP-Extra しか対応していません。(基本的に対応する予定もあ
 
 ## 変換方法
 
-[こちら](https://github.com/tuna2134/sbv2-api/tree/main/convert)を参照してください。
+[こちら](https://github.com/neodyland/sbv2-api/tree/main/scripts/convert)を参照してください。
 
 ## Todo
 
@@ -26,22 +41,27 @@ JP-Extra しか対応していません。(基本的に対応する予定もあ
 - [x] GPU 対応(CUDA)
 - [x] GPU 対応(DirectML)
 - [x] GPU 対応(CoreML)
-- [ ] WASM 変換(依存ライブラリの関係により現在は不可)
-- [ ] arm64のdockerサポート
+- [x] WASM 変換
+- [x] arm64のdockerサポート
+- [x] aivis形式のサポート
+- [ ] MeCabを利用する
 
 ## 構造説明
 
-- `sbv2_api` - 推論用 REST API
-- `sbv2_core` - 推論コア部分
-- `docker` - docker ビルドスクリプト
-- `convert` - onnx, sbv2フォーマットへの変換スクリプト
+- `crates/sbv2_api` - 推論用 REST API
+- `crates/sbv2_core` - 推論コア部分
+- `scripts/docker` - docker ビルドスクリプト
+- `scripts/convert` - onnx, sbv2フォーマットへの変換スクリプト
 
 ## プログラミングある程度できる人向けREST API起動方法
 
 ### models をインストール
 
-https://huggingface.co/googlefan/sbv2_onnx_models/tree/main
-の`tokenizer.json`,`debert.onnx`,`tsukuyomi.sbv2`を models フォルダに配置
+https://huggingface.co/neody/sbv2-api-assets/tree/main/deberta
+から`tokenizer.json`,`debert.onnx`
+https://huggingface.co/neody/sbv2-api-assets/tree/main/model
+から`tsukuyomi.sbv2`
+を models フォルダに配置
 
 ### .env ファイルの作成
 
@@ -55,7 +75,7 @@ CPUの場合は
 ```sh
 docker run -it --rm -p 3000:3000 --name sbv2 \
 -v ./models:/work/models --env-file .env \
-ghcr.io/tuna2134/sbv2-api:cpu
+ghcr.io/neodyland/sbv2-api:cpu
 ```
 
 <details>
@@ -70,7 +90,7 @@ CPUの場合は
 ```bash
 docker run --platform linux/amd64 -it --rm -p 3000:3000 --name sbv2 \
 -v ./models:/work/models --env-file .env \
-ghcr.io/tuna2134/sbv2-api:cpu
+ghcr.io/neodyland/sbv2-api:cpu
 ```
 </details>
 
@@ -79,7 +99,7 @@ CUDAの場合は
 docker run -it --rm -p 3000:3000 --name sbv2 \
 -v ./models:/work/models --env-file .env \
 --gpus all \
-ghcr.io/tuna2134/sbv2-api:cuda
+ghcr.io/neodyland/sbv2-api:cuda
 ```
 
 ### 起動確認
@@ -110,8 +130,10 @@ curl http://localhost:3000/models
 - `ADDR` `localhost:3000`などのようにサーバー起動アドレスをコントロールできます。
 - `MODELS_PATH` sbv2モデルの存在するフォルダを指定できます。
 - `RUST_LOG` おなじみlog levelです。
+- `HOLDER_MAX_LOADED_MODElS` RAMにロードされるモデルの最大数を指定します。
 
 ## 謝辞
 
-- [litagin02/Style-Bert-VITS2](https://github.com/litagin02/Style-Bert-VITS2) - このコードの書くにあたり、ベースとなる部分を参考にさせていただきました。
+- [litagin02/Style-Bert-VITS2](https://github.com/litagin02/Style-Bert-VITS2) - このコードを書くにあたり、ベースとなる部分を参考にさせていただきました。
 - [Googlefan](https://github.com/Googlefan256) - 彼にモデルを ONNX ヘ変換および効率化をする方法を教わりました。
+- [Aivis Project](https://github.com/Aivis-Project/AivisSpeech-Engine) - 辞書部分

content.txt

@@ -1 +0,0 @@
-10,000年前までコロナが流行っていました

convert/requirements.txt

@@ -1,5 +0,0 @@
-style-bert-vits2
-onnxsim
-numpy<2
-zstandard
-onnxruntime

crates/sbv2_api/Cargo.toml

@@ -0,0 +1,29 @@
+[package]
+name = "sbv2_api"
+version.workspace = true
+edition.workspace = true
+description.workspace = true
+readme.workspace = true
+repository.workspace = true
+documentation.workspace = true
+license.workspace = true
+
+[dependencies]
+anyhow.workspace = true
+axum = "0.8.0"
+dotenvy.workspace = true
+env_logger.workspace = true
+log = "0.4.22"
+sbv2_core = { version = "0.2.0-alpha6", path = "../sbv2_core", features = ["aivmx"] }
+serde = { version = "1.0.210", features = ["derive"] }
+tokio = { version = "1.47.1", features = ["full"] }
+utoipa = { version = "5.4.0", features = ["axum_extras"] }
+utoipa-scalar = { version = "0.3.0", features = ["axum"] }
+
+[features]
+coreml = ["sbv2_core/coreml"]
+cuda = ["sbv2_core/cuda"]
+cuda_tf32 = ["sbv2_core/cuda_tf32"]
+dynamic = ["sbv2_core/dynamic"]
+directml = ["sbv2_core/directml"]
+tensorrt = ["sbv2_core/tensorrt"]

crates/sbv2_api/build.rs

@@ -0,0 +1,5 @@
+fn main() {
+    if cfg!(feature = "coreml") {
+        println!("cargo:rustc-link-arg=-fapple-link-rtlib");
+    }
+}

crates/sbv2_api/src/main.rs

@@ -11,10 +11,23 @@ use std::env;
 use std::sync::Arc;
 use tokio::fs;
 use tokio::sync::Mutex;
+use utoipa::{OpenApi, ToSchema};
+use utoipa_scalar::{Scalar, Servable};
 
 mod error;
 use crate::error::AppResult;
 
+#[derive(OpenApi)]
+#[openapi(paths(models, synthesize), components(schemas(SynthesizeRequest)))]
+struct ApiDoc;
+
+#[utoipa::path(
+    get,
+    path = "/models",
+    responses(
+        (status = 200, description = "Return model list", body = Vec<String>),
+    )
+)]
 async fn models(State(state): State<AppState>) -> AppResult<impl IntoResponse> {
     Ok(Json(state.tts_model.lock().await.models()))
 }
@@ -27,16 +40,40 @@ fn length_default() -> f32 {
     1.0
 }
 
-#[derive(Deserialize)]
+fn style_id_default() -> i32 {
+    0
+}
+
+fn speaker_id_default() -> i64 {
+    0
+}
+
+#[derive(Deserialize, ToSchema)]
 struct SynthesizeRequest {
     text: String,
     ident: String,
     #[serde(default = "sdp_default")]
+    #[schema(example = 0.0_f32)]
     sdp_ratio: f32,
     #[serde(default = "length_default")]
+    #[schema(example = 1.0_f32)]
     length_scale: f32,
+    #[serde(default = "style_id_default")]
+    #[schema(example = 0_i32)]
+    style_id: i32,
+    #[serde(default = "speaker_id_default")]
+    #[schema(example = 0_i64)]
+    speaker_id: i64,
 }
 
+#[utoipa::path(
+    post,
+    path = "/synthesize",
+    request_body = SynthesizeRequest,
+    responses(
+        (status = 200, description = "Return audio/wav", body = Vec<u8>, content_type = "audio/wav")
+    )
+)]
 async fn synthesize(
     State(state): State<AppState>,
     Json(SynthesizeRequest {
@@ -44,15 +81,18 @@ async fn synthesize(
         ident,
         sdp_ratio,
         length_scale,
+        style_id,
+        speaker_id,
     }): Json<SynthesizeRequest>,
 ) -> AppResult<impl IntoResponse> {
     log::debug!("processing request: text={text}, ident={ident}, sdp_ratio={sdp_ratio}, length_scale={length_scale}");
     let buffer = {
-        let tts_model = state.tts_model.lock().await;
+        let mut tts_model = state.tts_model.lock().await;
         tts_model.easy_synthesize(
             &ident,
             &text,
-            0,
+            style_id,
+            speaker_id,
             SynthesizeOptions {
                 sdp_ratio,
                 length_scale,
@@ -73,6 +113,9 @@ impl AppState {
         let mut tts_model = TTSModelHolder::new(
             &fs::read(env::var("BERT_MODEL_PATH")?).await?,
             &fs::read(env::var("TOKENIZER_PATH")?).await?,
+            env::var("HOLDER_MAX_LOADED_MODElS")
+                .ok()
+                .and_then(|x| x.parse().ok()),
         )?;
         let models = env::var("MODELS_PATH").unwrap_or("models".to_string());
         let mut f = fs::read_dir(&models).await?;
@@ -101,6 +144,20 @@ impl AppState {
                     log::warn!("Error loading {entry}: {e}");
                 };
                 log::info!("Loaded: {entry}");
+            } else if name.ends_with(".aivmx") {
+                let entry = &name[..name.len() - 6];
+                log::info!("Try loading: {entry}");
+                let aivmx_bytes = match fs::read(format!("{models}/{entry}.aivmx")).await {
+                    Ok(b) => b,
+                    Err(e) => {
+                        log::warn!("Error loading aivmx bytes from file {entry}: {e}");
+                        continue;
+                    }
+                };
+                if let Err(e) = tts_model.load_aivmx(entry, aivmx_bytes) {
+                    log::error!("Error loading {entry}: {e}");
+                }
+                log::info!("Loaded: {entry}");
             }
         }
         for entry in entries {
@@ -139,7 +196,8 @@ async fn main() -> anyhow::Result<()> {
         .route("/", get(|| async { "Hello, World!" }))
         .route("/synthesize", post(synthesize))
         .route("/models", get(models))
-        .with_state(AppState::new().await?);
+        .with_state(AppState::new().await?)
+        .merge(Scalar::with_url("/docs", ApiDoc::openapi()));
     let addr = env::var("ADDR").unwrap_or("0.0.0.0:3000".to_string());
     let listener = tokio::net::TcpListener::bind(&addr).await?;
     log::info!("Listening on {addr}");

crates/sbv2_bindings/Cargo.toml

@@ -0,0 +1,24 @@
+[package]
+name = "sbv2_bindings"
+version.workspace = true
+edition.workspace = true
+description.workspace = true
+readme.workspace = true
+repository.workspace = true
+documentation.workspace = true
+license.workspace = true
+
+# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
+[lib]
+name = "sbv2_bindings"
+crate-type = ["cdylib"]
+
+[dependencies]
+anyhow.workspace = true
+ndarray.workspace = true
+pyo3 = { version = "0.25.1", features = ["anyhow"] }
+sbv2_core = { path = "../sbv2_core", features = ["std"], default-features = false }
+
+[features]
+agpl_dict = ["sbv2_core/agpl_dict"]
+default = ["agpl_dict"]

crates/sbv2_bindings/pyproject.toml

@@ -11,5 +11,7 @@ classifiers = [
     "Programming Language :: Python :: Implementation :: PyPy",
 ]
 dynamic = ["version"]
+
 [tool.maturin]
 features = ["pyo3/extension-module"]
+strip = true

crates/sbv2_bindings/src/sbv2.rs

@@ -1,6 +1,6 @@
 use pyo3::prelude::*;
 use pyo3::types::PyBytes;
-use sbv2_core::tts::{TTSModelHolder, SynthesizeOptions};
+use sbv2_core::tts::{SynthesizeOptions, TTSModelHolder};
 
 use crate::style::StyleVector;
 
@@ -23,10 +23,15 @@ pub struct TTSModel {
 
 #[pymethods]
 impl TTSModel {
+    #[pyo3(signature = (bert_model_bytes, tokenizer_bytes, max_loaded_models=None))]
     #[new]
-    fn new(bert_model_bytes: Vec<u8>, tokenizer_bytes: Vec<u8>) -> anyhow::Result<Self> {
+    fn new(
+        bert_model_bytes: Vec<u8>,
+        tokenizer_bytes: Vec<u8>,
+        max_loaded_models: Option<usize>,
+    ) -> anyhow::Result<Self> {
         Ok(Self {
-            model: TTSModelHolder::new(bert_model_bytes, tokenizer_bytes)?,
+            model: TTSModelHolder::new(bert_model_bytes, tokenizer_bytes, max_loaded_models)?,
         })
     }
 
@@ -38,10 +43,21 @@ impl TTSModel {
     ///     BERTモデルのパス
     /// tokenizer_path : str
     ///     トークナイザーのパス
+    /// max_loaded_models: int | None
+    ///     同時にVRAMに存在するモデルの数
+    #[pyo3(signature = (bert_model_path, tokenizer_path, max_loaded_models=None))]
     #[staticmethod]
-    fn from_path(bert_model_path: String, tokenizer_path: String) -> anyhow::Result<Self> {
+    fn from_path(
+        bert_model_path: String,
+        tokenizer_path: String,
+        max_loaded_models: Option<usize>,
+    ) -> anyhow::Result<Self> {
         Ok(Self {
-            model: TTSModelHolder::new(fs::read(bert_model_path)?, fs::read(tokenizer_path)?)?,
+            model: TTSModelHolder::new(
+                fs::read(bert_model_path)?,
+                fs::read(tokenizer_path)?,
+                max_loaded_models,
+            )?,
         })
     }
 
@@ -91,7 +107,7 @@ impl TTSModel {
     /// style_vector : StyleVector
     ///     スタイルベクトル
     fn get_style_vector(
-        &self,
+        &mut self,
         ident: String,
         style_id: i32,
         weight: f32,
@@ -120,25 +136,32 @@ impl TTSModel {
     /// -------
     /// voice_data : bytes
     ///     音声データ
+    #[allow(clippy::too_many_arguments)]
     fn synthesize<'p>(
-        &'p self,
+        &'p mut self,
         py: Python<'p>,
         text: String,
         ident: String,
         style_id: i32,
+        speaker_id: i64,
         sdp_ratio: f32,
         length_scale: f32,
-    ) -> anyhow::Result<Bound<PyBytes>> {
+    ) -> anyhow::Result<Bound<'p, PyBytes>> {
         let data = self.model.easy_synthesize(
             ident.as_str(),
             &text,
             style_id,
+            speaker_id,
             SynthesizeOptions {
                 sdp_ratio,
                 length_scale,
                 ..Default::default()
             },
         )?;
-        Ok(PyBytes::new_bound(py, &data))
+        Ok(PyBytes::new(py, &data))
+    }
+
+    fn unload(&mut self, ident: String) -> bool {
+        self.model.unload(ident)
     }
 }

crates/sbv2_core/Cargo.toml

@@ -0,0 +1,47 @@
+[package]
+name = "sbv2_core"
+version.workspace = true
+edition.workspace = true
+description.workspace = true
+readme.workspace = true
+repository.workspace = true
+documentation.workspace = true
+license.workspace = true
+
+[dependencies]
+anyhow.workspace = true
+base64 = { version = "0.22.1", optional = true }
+dotenvy.workspace = true
+env_logger.workspace = true
+hound = "3.5.1"
+jpreprocess = { version = "0.12.0", features = ["naist-jdic"] }
+ndarray.workspace = true
+npyz = { version = "0.8.4", optional = true }
+num_cpus = "1.17.0"
+once_cell.workspace = true
+ort = { git = "https://github.com/pykeio/ort.git", version = "2.0.0-rc.9", optional = true }
+regex = "1.10.6"
+serde = { version = "1.0.210", features = ["derive"] }
+serde_json = "1.0.142"
+tar = "0.4.41"
+thiserror = "2.0.15"
+tokenizers = { version = "0.21.4", default-features = false }
+zstd = "0.13.2"
+
+[features]
+cuda = ["ort/cuda", "std"]
+cuda_tf32 = ["std", "cuda"]
+agpl_dict = []
+std = ["dep:ort", "tokenizers/progressbar", "tokenizers/onig", "tokenizers/esaxx_fast"]
+dynamic = ["ort/load-dynamic", "std"]
+directml = ["ort/directml", "std"]
+tensorrt = ["ort/tensorrt", "std"]
+coreml = ["ort/coreml", "std"]
+default = ["std", "agpl_dict"]
+no_std = ["tokenizers/unstable_wasm"]
+aivmx = ["npyz", "base64"]
+base64 = ["dep:base64"]
+
+[build-dependencies]
+dirs = "6.0.0"
+ureq = "3.1.0"

crates/sbv2_core/build.rs

@@ -0,0 +1,31 @@
+use dirs::home_dir;
+use std::env;
+use std::fs;
+use std::io::copy;
+use std::path::PathBuf;
+
+fn main() -> Result<(), Box<dyn std::error::Error>> {
+    let static_dir = home_dir().unwrap().join(".cache/sbv2");
+    let static_path = static_dir.join("all.bin");
+    let out_path = PathBuf::from(&env::var("OUT_DIR").unwrap()).join("all.bin");
+    println!("cargo:rerun-if-changed=build.rs");
+    if static_path.exists() {
+        println!("cargo:info=Dictionary file already exists, skipping download.");
+    } else {
+        println!("cargo:warning=Downloading dictionary file...");
+        let mut response =
+            ureq::get("https://huggingface.co/neody/sbv2-api-assets/resolve/main/dic/all.bin")
+                .call()?;
+        let mut response = response.body_mut().as_reader();
+        if !static_dir.exists() {
+            fs::create_dir_all(static_dir)?;
+        }
+        let mut file = fs::File::create(&static_path)?;
+        copy(&mut response, &mut file)?;
+    }
+    if !out_path.exists() && fs::hard_link(&static_path, &out_path).is_err() {
+        println!("cargo:warning=Failed to create hard link, copying instead.");
+        fs::copy(static_path, out_path)?;
+    }
+    Ok(())
+}

crates/sbv2_core/src/bert.rs

@@ -0,0 +1,22 @@
+use crate::error::Result;
+use ndarray::{Array2, Ix2};
+use ort::session::Session;
+use ort::value::TensorRef;
+
+pub fn predict(
+    session: &mut Session,
+    token_ids: Vec<i64>,
+    attention_masks: Vec<i64>,
+) -> Result<Array2<f32>> {
+    let outputs = session.run(
+        ort::inputs! {
+            "input_ids" => TensorRef::from_array_view((vec![1, token_ids.len() as i64], token_ids.as_slice()))?,
+            "attention_mask" => TensorRef::from_array_view((vec![1, attention_masks.len() as i64], attention_masks.as_slice()))?,
+        }
+    )?;
+    let output = outputs["output"]
+        .try_extract_array::<f32>()?
+        .into_dimensionality::<Ix2>()?
+        .to_owned();
+    Ok(output)
+}

crates/sbv2_core/src/error.rs

@@ -6,6 +6,9 @@ pub enum Error {
     TokenizerError(#[from] tokenizers::Error),
     #[error("JPreprocess error: {0}")]
     JPreprocessError(#[from] jpreprocess::error::JPreprocessError),
+    #[error("Lindera error: {0}")]
+    LinderaError(String),
+    #[cfg(feature = "std")]
     #[error("ONNX error: {0}")]
     OrtError(#[from] ort::Error),
     #[error("NDArray error: {0}")]
@@ -20,6 +23,13 @@ pub enum Error {
     HoundError(#[from] hound::Error),
     #[error("model not found error")]
     ModelNotFoundError(String),
+    #[cfg(feature = "base64")]
+    #[error("base64 error")]
+    Base64Error(#[from] base64::DecodeError),
+    #[error("other")]
+    OtherError(String),
+    #[error("Style error: {0}")]
+    StyleError(String),
 }
 
 pub type Result<T> = std::result::Result<T, Error>;

crates/sbv2_core/src/jtalk.rs

@@ -0,0 +1,621 @@
+/*
+このファイルのコードは
+https://github.com/litagin02/Style-Bert-VITS2/blob/master/style_bert_vits2/nlp/japanese/g2p.py
+を参考にRustに書き換えています。
+
+以下はライセンスです。
+                   GNU LESSER GENERAL PUBLIC LICENSE
+                       Version 3, 29 June 2007
+
+ Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+
+  This version of the GNU Lesser General Public License incorporates
+the terms and conditions of version 3 of the GNU General Public
+License, supplemented by the additional permissions listed below.
+
+  0. Additional Definitions.
+
+  As used herein, "this License" refers to version 3 of the GNU Lesser
+General Public License, and the "GNU GPL" refers to version 3 of the GNU
+General Public License.
+
+  "The Library" refers to a covered work governed by this License,
+other than an Application or a Combined Work as defined below.
+
+  An "Application" is any work that makes use of an interface provided
+by the Library, but which is not otherwise based on the Library.
+Defining a subclass of a class defined by the Library is deemed a mode
+of using an interface provided by the Library.
+
+  A "Combined Work" is a work produced by combining or linking an
+Application with the Library.  The particular version of the Library
+with which the Combined Work was made is also called the "Linked
+Version".
+
+  The "Minimal Corresponding Source" for a Combined Work means the
+Corresponding Source for the Combined Work, excluding any source code
+for portions of the Combined Work that, considered in isolation, are
+based on the Application, and not on the Linked Version.
+
+  The "Corresponding Application Code" for a Combined Work means the
+object code and/or source code for the Application, including any data
+and utility programs needed for reproducing the Combined Work from the
+Application, but excluding the System Libraries of the Combined Work.
+
+  1. Exception to Section 3 of the GNU GPL.
+
+  You may convey a covered work under sections 3 and 4 of this License
+without being bound by section 3 of the GNU GPL.
+
+  2. Conveying Modified Versions.
+
+  If you modify a copy of the Library, and, in your modifications, a
+facility refers to a function or data to be supplied by an Application
+that uses the facility (other than as an argument passed when the
+facility is invoked), then you may convey a copy of the modified
+version:
+
+   a) under this License, provided that you make a good faith effort to
+   ensure that, in the event an Application does not supply the
+   function or data, the facility still operates, and performs
+   whatever part of its purpose remains meaningful, or
+
+   b) under the GNU GPL, with none of the additional permissions of
+   this License applicable to that copy.
+
+  3. Object Code Incorporating Material from Library Header Files.
+
+  The object code form of an Application may incorporate material from
+a header file that is part of the Library.  You may convey such object
+code under terms of your choice, provided that, if the incorporated
+material is not limited to numerical parameters, data structure
+layouts and accessors, or small macros, inline functions and templates
+(ten or fewer lines in length), you do both of the following:
+
+   a) Give prominent notice with each copy of the object code that the
+   Library is used in it and that the Library and its use are
+   covered by this License.
+
+   b) Accompany the object code with a copy of the GNU GPL and this license
+   document.
+
+  4. Combined Works.
+
+  You may convey a Combined Work under terms of your choice that,
+taken together, effectively do not restrict modification of the
+portions of the Library contained in the Combined Work and reverse
+engineering for debugging such modifications, if you also do each of
+the following:
+
+   a) Give prominent notice with each copy of the Combined Work that
+   the Library is used in it and that the Library and its use are
+   covered by this License.
+
+   b) Accompany the Combined Work with a copy of the GNU GPL and this license
+   document.
+
+   c) For a Combined Work that displays copyright notices during
+   execution, include the copyright notice for the Library among
+   these notices, as well as a reference directing the user to the
+   copies of the GNU GPL and this license document.
+
+   d) Do one of the following:
+
+       0) Convey the Minimal Corresponding Source under the terms of this
+       License, and the Corresponding Application Code in a form
+       suitable for, and under terms that permit, the user to
+       recombine or relink the Application with a modified version of
+       the Linked Version to produce a modified Combined Work, in the
+       manner specified by section 6 of the GNU GPL for conveying
+       Corresponding Source.
+
+       1) Use a suitable shared library mechanism for linking with the
+       Library.  A suitable mechanism is one that (a) uses at run time
+       a copy of the Library already present on the user's computer
+       system, and (b) will operate properly with a modified version
+       of the Library that is interface-compatible with the Linked
+       Version.
+
+   e) Provide Installation Information, but only if you would otherwise
+   be required to provide such information under section 6 of the
+   GNU GPL, and only to the extent that such information is
+   necessary to install and execute a modified version of the
+   Combined Work produced by recombining or relinking the
+   Application with a modified version of the Linked Version. (If
+   you use option 4d0, the Installation Information must accompany
+   the Minimal Corresponding Source and Corresponding Application
+   Code. If you use option 4d1, you must provide the Installation
+   Information in the manner specified by section 6 of the GNU GPL
+   for conveying Corresponding Source.)
+
+  5. Combined Libraries.
+
+  You may place library facilities that are a work based on the
+Library side by side in a single library together with other library
+facilities that are not Applications and are not covered by this
+License, and convey such a combined library under terms of your
+choice, if you do both of the following:
+
+   a) Accompany the combined library with a copy of the same work based
+   on the Library, uncombined with any other library facilities,
+   conveyed under the terms of this License.
+
+   b) Give prominent notice with the combined library that part of it
+   is a work based on the Library, and explaining where to find the
+   accompanying uncombined form of the same work.
+
+  6. Revised Versions of the GNU Lesser General Public License.
+
+  The Free Software Foundation may publish revised and/or new versions
+of the GNU Lesser General Public License from time to time. Such new
+versions will be similar in spirit to the present version, but may
+differ in detail to address new problems or concerns.
+
+  Each version is given a distinguishing version number. If the
+Library as you received it specifies that a certain numbered version
+of the GNU Lesser General Public License "or any later version"
+applies to it, you have the option of following the terms and
+conditions either of that published version or of any later version
+published by the Free Software Foundation. If the Library as you
+received it does not specify a version number of the GNU Lesser
+General Public License, you may choose any version of the GNU Lesser
+General Public License ever published by the Free Software Foundation.
+
+  If the Library as you received it specifies that a proxy can decide
+whether future versions of the GNU Lesser General Public License shall
+apply, that proxy's public statement of acceptance of any version is
+permanent authorization for you to choose that version for the
+Library.
+*/
+use crate::error::{Error, Result};
+use crate::mora::{CONSONANTS, MORA_KATA_TO_MORA_PHONEMES, MORA_PHONEMES_TO_MORA_KATA, VOWELS};
+use crate::norm::{replace_punctuation, PUNCTUATIONS};
+use jpreprocess::{kind, DefaultTokenizer, JPreprocess, SystemDictionaryConfig, UserDictionary};
+use once_cell::sync::Lazy;
+use regex::Regex;
+use std::cmp::Reverse;
+use std::collections::HashSet;
+use std::sync::Arc;
+
+type JPreprocessType = JPreprocess<DefaultTokenizer>;
+
+#[cfg(feature = "agpl_dict")]
+fn agpl_dict() -> Result<Option<UserDictionary>> {
+    Ok(Some(
+        UserDictionary::load(include_bytes!(concat!(env!("OUT_DIR"), "/all.bin")))
+            .map_err(|e| Error::LinderaError(e.to_string()))?,
+    ))
+}
+
+#[cfg(not(feature = "agpl_dict"))]
+fn agpl_dict() -> Result<Option<UserDictionary>> {
+    Ok(None)
+}
+
+fn initialize_jtalk() -> Result<JPreprocessType> {
+    let sdic =
+        SystemDictionaryConfig::Bundled(kind::JPreprocessDictionaryKind::NaistJdic).load()?;
+    let jpreprocess = JPreprocess::with_dictionaries(sdic, agpl_dict()?);
+    Ok(jpreprocess)
+}
+
+macro_rules! hash_set {
+    ($($elem:expr),* $(,)?) => {{
+        let mut set = HashSet::new();
+        $(
+            set.insert($elem);
+        )*
+        set
+    }};
+}
+
+pub struct JTalk {
+    pub jpreprocess: Arc<JPreprocessType>,
+}
+
+impl JTalk {
+    pub fn new() -> Result<Self> {
+        let jpreprocess = Arc::new(initialize_jtalk()?);
+        Ok(Self { jpreprocess })
+    }
+
+    pub fn num2word(&self, text: &str) -> Result<String> {
+        let mut parsed = self.jpreprocess.text_to_njd(text)?;
+        parsed.preprocess();
+        let texts: Vec<String> = parsed
+            .nodes
+            .iter()
+            .map(|x| x.get_string().to_string())
+            .collect();
+        Ok(texts.join(""))
+    }
+
+    pub fn process_text(&self, text: &str) -> Result<JTalkProcess> {
+        let parsed = self.jpreprocess.run_frontend(text)?;
+        let jtalk_process = JTalkProcess::new(Arc::clone(&self.jpreprocess), parsed);
+        Ok(jtalk_process)
+    }
+}
+
+static KATAKANA_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"[\u30A0-\u30FF]+").unwrap());
+static MORA_PATTERN: Lazy<Vec<String>> = Lazy::new(|| {
+    let mut sorted_keys: Vec<String> = MORA_KATA_TO_MORA_PHONEMES.keys().cloned().collect();
+    sorted_keys.sort_by_key(|b| Reverse(b.len()));
+    sorted_keys
+});
+static LONG_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\w)(ー*)").unwrap());
+
+fn phone_tone_to_kana(phones: Vec<String>, tones: Vec<i32>) -> Vec<(String, i32)> {
+    let phones = &phones[1..];
+    let tones = &tones[1..];
+    let mut results = Vec::new();
+    let mut current_mora = String::new();
+    for ((phone, _next_phone), (&tone, &next_tone)) in phones
+        .iter()
+        .zip(phones.iter().skip(1))
+        .zip(tones.iter().zip(tones.iter().skip(1)))
+    {
+        if PUNCTUATIONS.contains(&phone.clone().as_str()) {
+            results.push((phone.to_string(), tone));
+            continue;
+        }
+        if CONSONANTS.contains(&phone.clone()) {
+            assert_eq!(current_mora, "");
+            assert_eq!(tone, next_tone);
+            current_mora = phone.to_string()
+        } else {
+            current_mora += phone;
+            let kana = MORA_PHONEMES_TO_MORA_KATA.get(&current_mora).unwrap();
+            results.push((kana.to_string(), tone));
+            current_mora = String::new();
+        }
+    }
+    results
+}
+
+pub struct JTalkProcess {
+    jpreprocess: Arc<JPreprocessType>,
+    parsed: Vec<String>,
+}
+
+impl JTalkProcess {
+    fn new(jpreprocess: Arc<JPreprocessType>, parsed: Vec<String>) -> Self {
+        Self {
+            jpreprocess,
+            parsed,
+        }
+    }
+
+    fn fix_phone_tone(&self, phone_tone_list: Vec<(String, i32)>) -> Result<Vec<(String, i32)>> {
+        let tone_values: HashSet<i32> = phone_tone_list
+            .iter()
+            .map(|(_letter, tone)| *tone)
+            .collect();
+        if tone_values.len() == 1 {
+            assert!(tone_values == hash_set![0], "{tone_values:?}");
+            Ok(phone_tone_list)
+        } else if tone_values.len() == 2 {
+            if tone_values == hash_set![0, 1] {
+                Ok(phone_tone_list)
+            } else if tone_values == hash_set![-1, 0] {
+                Ok(phone_tone_list
+                    .iter()
+                    .map(|x| {
+                        let new_tone = if x.1 == -1 { 0 } else { 1 };
+                        (x.0.clone(), new_tone)
+                    })
+                    .collect())
+            } else {
+                Err(Error::ValueError("Invalid tone values 0".to_string()))
+            }
+        } else {
+            Err(Error::ValueError("Invalid tone values 1".to_string()))
+        }
+    }
+
+    pub fn g2p(&self) -> Result<(Vec<String>, Vec<i32>, Vec<i32>)> {
+        let phone_tone_list_wo_punct = self.g2phone_tone_wo_punct()?;
+        let (seq_text, seq_kata) = self.text_to_seq_kata()?;
+        let sep_phonemes = JTalkProcess::handle_long(
+            seq_kata
+                .iter()
+                .map(|x| JTalkProcess::kata_to_phoneme_list(x.clone()).unwrap())
+                .collect(),
+        );
+        let phone_w_punct: Vec<String> = sep_phonemes
+            .iter()
+            .flat_map(|x| x.iter())
+            .cloned()
+            .collect();
+
+        let mut phone_tone_list =
+            JTalkProcess::align_tones(phone_w_punct, phone_tone_list_wo_punct)?;
+
+        let mut sep_tokenized: Vec<Vec<String>> = Vec::new();
+        for seq_text_item in &seq_text {
+            let text = seq_text_item.clone();
+            if !PUNCTUATIONS.contains(&text.as_str()) {
+                sep_tokenized.push(text.chars().map(|x| x.to_string()).collect());
+            } else {
+                sep_tokenized.push(vec![text]);
+            }
+        }
+
+        let mut word2ph = Vec::new();
+        for (token, phoneme) in sep_tokenized.iter().zip(sep_phonemes.iter()) {
+            let phone_len = phoneme.len() as i32;
+            let word_len = token.len() as i32;
+            word2ph.append(&mut JTalkProcess::distribute_phone(phone_len, word_len));
+        }
+
+        let mut new_phone_tone_list = vec![("_".to_string(), 0)];
+        new_phone_tone_list.append(&mut phone_tone_list);
+        new_phone_tone_list.push(("_".to_string(), 0));
+
+        let mut new_word2ph = vec![1];
+        new_word2ph.extend(word2ph.clone());
+        new_word2ph.push(1);
+
+        let phones: Vec<String> = new_phone_tone_list.iter().map(|(x, _)| x.clone()).collect();
+        let tones: Vec<i32> = new_phone_tone_list.iter().map(|(_, x)| *x).collect();
+
+        Ok((phones, tones, new_word2ph))
+    }
+
+    pub fn g2kana_tone(&self) -> Result<Vec<(String, i32)>> {
+        let (phones, tones, _) = self.g2p()?;
+        Ok(phone_tone_to_kana(phones, tones))
+    }
+
+    fn distribute_phone(n_phone: i32, n_word: i32) -> Vec<i32> {
+        let mut phones_per_word = vec![0; n_word as usize];
+        for _ in 0..n_phone {
+            let min_task = phones_per_word.iter().min().unwrap();
+            let min_index = phones_per_word
+                .iter()
+                .position(|&x| x == *min_task)
+                .unwrap();
+            phones_per_word[min_index] += 1;
+        }
+        phones_per_word
+    }
+
+    fn align_tones(
+        phone_with_punct: Vec<String>,
+        phone_tone_list: Vec<(String, i32)>,
+    ) -> Result<Vec<(String, i32)>> {
+        let mut result: Vec<(String, i32)> = Vec::new();
+        let mut tone_index = 0;
+        for phone in phone_with_punct.clone() {
+            if tone_index >= phone_tone_list.len() {
+                result.push((phone, 0));
+            } else if phone == phone_tone_list[tone_index].0 {
+                result.push((phone, phone_tone_list[tone_index].1));
+                tone_index += 1;
+            } else if PUNCTUATIONS.contains(&phone.as_str()) {
+                result.push((phone, 0));
+            } else {
+                println!("phones {phone_with_punct:?}");
+                println!("phone_tone_list: {phone_tone_list:?}");
+                println!("result: {result:?}");
+                println!("tone_index: {tone_index:?}");
+                println!("phone: {phone:?}");
+                return Err(Error::ValueError(format!("Mismatched phoneme: {phone}")));
+            }
+        }
+
+        Ok(result)
+    }
+
+    fn handle_long(mut sep_phonemes: Vec<Vec<String>>) -> Vec<Vec<String>> {
+        for i in 0..sep_phonemes.len() {
+            if sep_phonemes[i].is_empty() {
+                continue;
+            }
+            if sep_phonemes[i][0] == "ー" {
+                if i != 0 {
+                    let prev_phoneme = sep_phonemes[i - 1].last().unwrap();
+                    if VOWELS.contains(&prev_phoneme.as_str()) {
+                        sep_phonemes[i][0] = prev_phoneme.clone();
+                    } else {
+                        sep_phonemes[i][0] = "ー".to_string();
+                    }
+                } else {
+                    sep_phonemes[i][0] = "ー".to_string();
+                }
+            }
+            if sep_phonemes[i].contains(&"ー".to_string()) {
+                for e in 0..sep_phonemes[i].len() {
+                    if sep_phonemes[i][e] == "ー" {
+                        sep_phonemes[i][e] =
+                            sep_phonemes[i][e - 1].chars().last().unwrap().to_string();
+                    }
+                }
+            }
+        }
+        sep_phonemes
+    }
+
+    fn kata_to_phoneme_list(mut text: String) -> Result<Vec<String>> {
+        let chars: HashSet<String> = text.chars().map(|x| x.to_string()).collect();
+        if chars.is_subset(&HashSet::from_iter(
+            PUNCTUATIONS.iter().map(|x| x.to_string()),
+        )) {
+            return Ok(text.chars().map(|x| x.to_string()).collect());
+        }
+        if !KATAKANA_PATTERN.is_match(&text) {
+            return Err(Error::ValueError(format!(
+                "Input must be katakana only: {text}"
+            )));
+        }
+
+        for mora in MORA_PATTERN.iter() {
+            let mora = mora.to_string();
+            let (consonant, vowel) = MORA_KATA_TO_MORA_PHONEMES.get(&mora).unwrap();
+            if consonant.is_none() {
+                text = text.replace(&mora, &format!(" {vowel}"));
+            } else {
+                text = text.replace(
+                    &mora,
+                    &format!(" {} {}", consonant.as_ref().unwrap(), vowel),
+                );
+            }
+        }
+
+        let long_replacement = |m: &regex::Captures| {
+            let result = m.get(1).unwrap().as_str().to_string();
+            let mut second = String::new();
+            for _ in 0..m.get(2).unwrap().as_str().char_indices().count() {
+                second += &format!(" {}", m.get(1).unwrap().as_str());
+            }
+            result + &second
+        };
+        text = LONG_PATTERN
+            .replace_all(&text, long_replacement)
+            .to_string();
+
+        let data = text.trim().split(' ').map(|x| x.to_string()).collect();
+
+        Ok(data)
+    }
+
+    pub fn text_to_seq_kata(&self) -> Result<(Vec<String>, Vec<String>)> {
+        let mut seq_kata = vec![];
+        let mut seq_text = vec![];
+
+        for parts in &self.parsed {
+            let (string, pron) = self.parse_to_string_and_pron(parts.clone());
+            let mut yomi = pron.replace('’', "");
+            let word = replace_punctuation(string);
+            assert!(!yomi.is_empty(), "Empty yomi: {word}");
+            if yomi == "、" {
+                if !word
+                    .chars()
+                    .all(|x| PUNCTUATIONS.contains(&x.to_string().as_str()))
+                {
+                    yomi = "'".repeat(word.len());
+                } else {
+                    yomi = word.clone();
+                }
+            } else if yomi == "？" {
+                assert!(word == "?", "yomi `？` comes from: {word}");
+                yomi = "?".to_string();
+            }
+            seq_text.push(word);
+            seq_kata.push(yomi);
+        }
+        Ok((seq_text, seq_kata))
+    }
+
+    fn parse_to_string_and_pron(&self, parts: String) -> (String, String) {
+        let part_lists: Vec<String> = parts.split(',').map(|x| x.to_string()).collect();
+        (part_lists[0].clone(), part_lists[9].clone())
+    }
+
+    fn g2phone_tone_wo_punct(&self) -> Result<Vec<(String, i32)>> {
+        let prosodies = self.g2p_prosody()?;
+
+        let mut results: Vec<(String, i32)> = Vec::new();
+        let mut current_phrase: Vec<(String, i32)> = Vec::new();
+        let mut current_tone = 0;
+
+        for (i, letter) in prosodies.iter().enumerate() {
+            if letter == "^" {
+                assert!(i == 0);
+            } else if ["$", "?", "_", "#"].contains(&letter.as_str()) {
+                results.extend(self.fix_phone_tone(current_phrase.clone())?);
+                if ["$", "?"].contains(&letter.as_str()) {
+                    assert!(i == prosodies.len() - 1);
+                }
+                current_phrase = Vec::new();
+                current_tone = 0;
+            } else if letter == "[" {
+                current_tone += 1;
+            } else if letter == "]" {
+                current_tone -= 1;
+            } else {
+                let new_letter = if letter == "cl" {
+                    "q".to_string()
+                } else {
+                    letter.clone()
+                };
+                current_phrase.push((new_letter, current_tone));
+            }
+        }
+
+        Ok(results)
+    }
+
+    fn g2p_prosody(&self) -> Result<Vec<String>> {
+        let labels = self.jpreprocess.make_label(self.parsed.clone());
+
+        let mut phones: Vec<String> = Vec::new();
+        for (i, label) in labels.iter().enumerate() {
+            let mut p3 = label.phoneme.c.clone().unwrap();
+            if "AIUEO".contains(&p3) {
+                // 文字をlowerする
+                p3 = p3.to_lowercase();
+            }
+            if p3 == "sil" {
+                assert!(i == 0 || i == labels.len() - 1);
+                if i == 0 {
+                    phones.push("^".to_string());
+                } else if i == labels.len() - 1 {
+                    let e3 = label.accent_phrase_prev.clone().unwrap().is_interrogative;
+                    if e3 {
+                        phones.push("$".to_string());
+                    } else {
+                        phones.push("?".to_string());
+                    }
+                }
+                continue;
+            } else if p3 == "pau" {
+                phones.push("_".to_string());
+                continue;
+            } else {
+                phones.push(p3.clone());
+            }
+
+            let a1 = if let Some(mora) = &label.mora {
+                mora.relative_accent_position as i32
+            } else {
+                -50
+            };
+            let a2 = if let Some(mora) = &label.mora {
+                mora.position_forward as i32
+            } else {
+                -50
+            };
+            let a3 = if let Some(mora) = &label.mora {
+                mora.position_backward as i32
+            } else {
+                -50
+            };
+
+            let f1 = if let Some(accent_phrase) = &label.accent_phrase_curr {
+                accent_phrase.mora_count as i32
+            } else {
+                -50
+            };
+
+            let a2_next = if let Some(mora) = &labels[i + 1].mora {
+                mora.position_forward as i32
+            } else {
+                -50
+            };
+
+            if a3 == 1 && a2_next == 1 && "aeiouAEIOUNcl".contains(&p3) {
+                phones.push("#".to_string());
+            } else if a1 == 0 && a2_next == a2 + 1 && a2 != f1 {
+                phones.push("]".to_string());
+            } else if a2 == 1 && a2_next == 2 {
+                phones.push("[".to_string());
+            }
+        }
+
+        Ok(phones)
+    }
+}

crates/sbv2_core/src/lib.rs

@@ -1,11 +1,16 @@
+#[cfg(feature = "std")]
 pub mod bert;
 pub mod error;
 pub mod jtalk;
+#[cfg(feature = "std")]
 pub mod model;
 pub mod mora;
 pub mod nlp;
 pub mod norm;
+pub mod sbv2file;
 pub mod style;
 pub mod tokenizer;
+#[cfg(feature = "std")]
 pub mod tts;
+pub mod tts_util;
 pub mod utils;

crates/sbv2_core/src/main.rs

@@ -0,0 +1,48 @@
+use std::env;
+use std::fs;
+
+#[cfg(feature = "std")]
+fn main_inner() -> anyhow::Result<()> {
+    use sbv2_core::tts;
+    dotenvy::dotenv_override().ok();
+    env_logger::init();
+    let text = "今日の天気は快晴です。";
+    let ident = "aaa";
+    let mut tts_holder = tts::TTSModelHolder::new(
+        &fs::read(env::var("BERT_MODEL_PATH")?)?,
+        &fs::read(env::var("TOKENIZER_PATH")?)?,
+        env::var("HOLDER_MAX_LOADED_MODElS")
+            .ok()
+            .and_then(|x| x.parse().ok()),
+    )?;
+    let mp = env::var("MODEL_PATH")?;
+    let b = fs::read(&mp)?;
+    #[cfg(not(feature = "aivmx"))]
+    {
+        tts_holder.load_sbv2file(ident, b)?;
+    }
+    #[cfg(feature = "aivmx")]
+    {
+        if mp.ends_with(".sbv2") {
+            tts_holder.load_sbv2file(ident, b)?;
+        } else {
+            tts_holder.load_aivmx(ident, b)?;
+        }
+    }
+
+    let audio = tts_holder.easy_synthesize(ident, text, 0, 0, tts::SynthesizeOptions::default())?;
+    fs::write("output.wav", audio)?;
+
+    Ok(())
+}
+
+#[cfg(not(feature = "std"))]
+fn main_inner() -> anyhow::Result<()> {
+    Ok(())
+}
+
+fn main() {
+    if let Err(e) = main_inner() {
+        println!("Error: {e}");
+    }
+}

crates/sbv2_core/src/model.rs

@@ -0,0 +1,106 @@
+use crate::error::Result;
+use ndarray::{array, Array1, Array2, Array3, Axis, Ix3};
+use ort::session::{builder::GraphOptimizationLevel, Session};
+
+#[allow(clippy::vec_init_then_push, unused_variables)]
+pub fn load_model<P: AsRef<[u8]>>(model_file: P, bert: bool) -> Result<Session> {
+    let mut exp = Vec::new();
+    #[cfg(feature = "tensorrt")]
+    {
+        if bert {
+            exp.push(
+                ort::execution_providers::TensorRTExecutionProvider::default()
+                    .with_fp16(true)
+                    .with_profile_min_shapes("input_ids:1x1,attention_mask:1x1")
+                    .with_profile_max_shapes("input_ids:1x100,attention_mask:1x100")
+                    .with_profile_opt_shapes("input_ids:1x25,attention_mask:1x25")
+                    .build(),
+            );
+        }
+    }
+    #[cfg(feature = "cuda")]
+    {
+        #[allow(unused_mut)]
+        let mut cuda = ort::execution_providers::CUDAExecutionProvider::default();
+        #[cfg(feature = "cuda_tf32")]
+        {
+            cuda = cuda.with_tf32(true);
+        }
+        exp.push(cuda.build());
+    }
+    #[cfg(feature = "directml")]
+    {
+        exp.push(ort::execution_providers::DirectMLExecutionProvider::default().build());
+    }
+    #[cfg(feature = "coreml")]
+    {
+        exp.push(ort::execution_providers::CoreMLExecutionProvider::default().build());
+    }
+    exp.push(ort::execution_providers::CPUExecutionProvider::default().build());
+    Ok(Session::builder()?
+        .with_execution_providers(exp)?
+        .with_optimization_level(GraphOptimizationLevel::Level3)?
+        .with_intra_threads(num_cpus::get_physical())?
+        .with_parallel_execution(true)?
+        .with_inter_threads(num_cpus::get_physical())?
+        .commit_from_memory(model_file.as_ref())?)
+}
+
+#[allow(clippy::too_many_arguments)]
+pub fn synthesize(
+    session: &mut Session,
+    bert_ori: Array2<f32>,
+    x_tst: Array1<i64>,
+    mut spk_ids: Array1<i64>,
+    tones: Array1<i64>,
+    lang_ids: Array1<i64>,
+    style_vector: Array1<f32>,
+    sdp_ratio: f32,
+    length_scale: f32,
+    noise_scale: f32,
+    noise_scale_w: f32,
+) -> Result<Array3<f32>> {
+    let bert_ori = bert_ori.insert_axis(Axis(0));
+    let bert_ori = bert_ori.as_standard_layout();
+    let bert = ort::value::TensorRef::from_array_view(&bert_ori)?;
+    let mut x_tst_lengths = array![x_tst.shape()[0] as i64];
+    let x_tst_lengths = ort::value::TensorRef::from_array_view(&mut x_tst_lengths)?;
+    let mut x_tst = x_tst.insert_axis(Axis(0));
+    let x_tst = ort::value::TensorRef::from_array_view(&mut x_tst)?;
+    let mut lang_ids = lang_ids.insert_axis(Axis(0));
+    let lang_ids = ort::value::TensorRef::from_array_view(&mut lang_ids)?;
+    let mut tones = tones.insert_axis(Axis(0));
+    let tones = ort::value::TensorRef::from_array_view(&mut tones)?;
+    let mut style_vector = style_vector.insert_axis(Axis(0));
+    let style_vector = ort::value::TensorRef::from_array_view(&mut style_vector)?;
+    let sid = ort::value::TensorRef::from_array_view(&mut spk_ids)?;
+    let sdp_ratio = vec![sdp_ratio];
+    let sdp_ratio = ort::value::TensorRef::from_array_view((vec![1_i64], sdp_ratio.as_slice()))?;
+    let length_scale = vec![length_scale];
+    let length_scale =
+        ort::value::TensorRef::from_array_view((vec![1_i64], length_scale.as_slice()))?;
+    let noise_scale = vec![noise_scale];
+    let noise_scale =
+        ort::value::TensorRef::from_array_view((vec![1_i64], noise_scale.as_slice()))?;
+    let noise_scale_w = vec![noise_scale_w];
+    let noise_scale_w =
+        ort::value::TensorRef::from_array_view((vec![1_i64], noise_scale_w.as_slice()))?;
+    let outputs = session.run(ort::inputs! {
+        "x_tst" =>  x_tst,
+        "x_tst_lengths" => x_tst_lengths,
+        "sid" => sid,
+        "tones" => tones,
+        "language" => lang_ids,
+        "bert" => bert,
+        "style_vec" => style_vector,
+        "sdp_ratio" => sdp_ratio,
+        "length_scale" => length_scale,
+        "noise_scale" => noise_scale,
+        "noise_scale_w" => noise_scale_w,
+    })?;
+    let audio_array = outputs["output"]
+        .try_extract_array::<f32>()?
+        .into_dimensionality::<Ix3>()?
+        .to_owned();
+    Ok(audio_array)
+}

crates/sbv2_core/src/mora.rs

@@ -25,6 +25,21 @@ static MORA_LIST_ADDITIONAL: Lazy<Vec<Mora>> = Lazy::new(|| {
     data.additional
 });
 
+pub static MORA_PHONEMES_TO_MORA_KATA: Lazy<HashMap<String, String>> = Lazy::new(|| {
+    let mut map = HashMap::new();
+    for mora in MORA_LIST_MINIMUM.iter() {
+        map.insert(
+            format!(
+                "{}{}",
+                mora.consonant.clone().unwrap_or("".to_string()),
+                mora.vowel
+            ),
+            mora.mora.clone(),
+        );
+    }
+    map
+});
+
 pub static MORA_KATA_TO_MORA_PHONEMES: Lazy<HashMap<String, (Option<String>, String)>> =
     Lazy::new(|| {
         let mut map = HashMap::new();
@@ -37,4 +52,12 @@ pub static MORA_KATA_TO_MORA_PHONEMES: Lazy<HashMap<String, (Option<String>, Str
         map
     });
 
+pub static CONSONANTS: Lazy<Vec<String>> = Lazy::new(|| {
+    let consonants = MORA_KATA_TO_MORA_PHONEMES
+        .values()
+        .filter_map(|(consonant, _)| consonant.clone())
+        .collect::<Vec<_>>();
+    consonants
+});
+
 pub const VOWELS: [&str; 6] = ["a", "i", "u", "e", "o", "N"];

crates/sbv2_core/src/sbv2file.rs

@@ -0,0 +1,37 @@
+use std::io::{Cursor, Read};
+
+use tar::Archive;
+use zstd::decode_all;
+
+use crate::error::{Error, Result};
+
+/// Parse a .sbv2 file binary
+///
+/// # Examples
+///
+/// ```rs
+/// parse_sbv2file("tsukuyomi", std::fs::read("tsukuyomi.sbv2")?)?;
+/// ```
+pub fn parse_sbv2file<P: AsRef<[u8]>>(sbv2_bytes: P) -> Result<(Vec<u8>, Vec<u8>)> {
+    let mut arc = Archive::new(Cursor::new(decode_all(Cursor::new(sbv2_bytes.as_ref()))?));
+    let mut vits2 = None;
+    let mut style_vectors = None;
+    let mut et = arc.entries()?;
+    while let Some(Ok(mut e)) = et.next() {
+        let pth = String::from_utf8_lossy(&e.path_bytes()).to_string();
+        let mut b = Vec::with_capacity(e.size() as usize);
+        e.read_to_end(&mut b)?;
+        match pth.as_str() {
+            "model.onnx" => vits2 = Some(b),
+            "style_vectors.json" => style_vectors = Some(b),
+            _ => continue,
+        }
+    }
+    if style_vectors.is_none() {
+        return Err(Error::ModelNotFoundError("style_vectors".to_string()));
+    }
+    if vits2.is_none() {
+        return Err(Error::ModelNotFoundError("vits2".to_string()));
+    }
+    Ok((style_vectors.unwrap(), vits2.unwrap()))
+}

crates/sbv2_core/src/style.rs

@@ -1,4 +1,4 @@
-use crate::error::Result;
+use crate::error::{Error, Result};
 use ndarray::{s, Array1, Array2};
 use serde::Deserialize;
 
@@ -21,6 +21,18 @@ pub fn get_style_vector(
     style_id: i32,
     weight: f32,
 ) -> Result<Array1<f32>> {
+    if style_vectors.shape().len() != 2 {
+        return Err(Error::StyleError(
+            "Invalid shape for style vectors".to_string(),
+        ));
+    }
+    if style_id < 0 || style_id >= style_vectors.shape()[0] as i32 {
+        return Err(Error::StyleError(format!(
+            "Invalid style ID: {}. Max ID: {}",
+            style_id,
+            style_vectors.shape()[0] - 1
+        )));
+    }
     let mean = style_vectors.slice(s![0, ..]).to_owned();
     let style_vector = style_vectors.slice(s![style_id as usize, ..]).to_owned();
     let diff = (style_vector - &mean) * weight;

crates/sbv2_core/src/tokenizer.rs

@@ -1,5 +1,5 @@
 use crate::error::Result;
-use tokenizers::Tokenizer;
+pub use tokenizers::Tokenizer;
 
 pub fn get_tokenizer<P: AsRef<[u8]>>(p: P) -> Result<Tokenizer> {
     let tokenizer = Tokenizer::from_bytes(p)?;

crates/sbv2_core/src/tts.rs

@@ -0,0 +1,470 @@
+use crate::error::{Error, Result};
+use crate::{jtalk, model, style, tokenizer, tts_util};
+#[cfg(feature = "aivmx")]
+use base64::prelude::{Engine as _, BASE64_STANDARD};
+#[cfg(feature = "aivmx")]
+use ndarray::ShapeBuilder;
+use ndarray::{concatenate, Array1, Array2, Array3, Axis};
+use ort::session::Session;
+#[cfg(feature = "aivmx")]
+use std::io::Cursor;
+use tokenizers::Tokenizer;
+
+#[derive(PartialEq, Eq, Clone)]
+pub struct TTSIdent(String);
+
+impl std::fmt::Display for TTSIdent {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        f.write_str(&self.0)?;
+        Ok(())
+    }
+}
+
+impl<S> From<S> for TTSIdent
+where
+    S: AsRef<str>,
+{
+    fn from(value: S) -> Self {
+        TTSIdent(value.as_ref().to_string())
+    }
+}
+
+pub struct TTSModel {
+    vits2: Option<Session>,
+    style_vectors: Array2<f32>,
+    ident: TTSIdent,
+    bytes: Option<Vec<u8>>,
+}
+
+/// High-level Style-Bert-VITS2's API
+pub struct TTSModelHolder {
+    tokenizer: Tokenizer,
+    bert: Session,
+    models: Vec<TTSModel>,
+    pub jtalk: jtalk::JTalk,
+    max_loaded_models: Option<usize>,
+}
+
+impl TTSModelHolder {
+    /// Initialize a new TTSModelHolder
+    ///
+    /// # Examples
+    ///
+    /// ```rs
+    /// let mut tts_holder = TTSModelHolder::new(std::fs::read("deberta.onnx")?, std::fs::read("tokenizer.json")?, None)?;
+    /// ```
+    pub fn new<P: AsRef<[u8]>>(
+        bert_model_bytes: P,
+        tokenizer_bytes: P,
+        max_loaded_models: Option<usize>,
+    ) -> Result<Self> {
+        let bert = model::load_model(bert_model_bytes, true)?;
+        let jtalk = jtalk::JTalk::new()?;
+        let tokenizer = tokenizer::get_tokenizer(tokenizer_bytes)?;
+        Ok(TTSModelHolder {
+            bert,
+            models: vec![],
+            jtalk,
+            tokenizer,
+            max_loaded_models,
+        })
+    }
+
+    /// Return a list of model names
+    pub fn models(&self) -> Vec<String> {
+        self.models.iter().map(|m| m.ident.to_string()).collect()
+    }
+
+    #[cfg(feature = "aivmx")]
+    pub fn load_aivmx<I: Into<TTSIdent>, P: AsRef<[u8]>>(
+        &mut self,
+        ident: I,
+        aivmx_bytes: P,
+    ) -> Result<()> {
+        let ident = ident.into();
+        if self.find_model(ident.clone()).is_err() {
+            let mut load = true;
+            if let Some(max) = self.max_loaded_models {
+                if self.models.iter().filter(|x| x.vits2.is_some()).count() >= max {
+                    load = false;
+                }
+            }
+            let model = model::load_model(&aivmx_bytes, false)?;
+            let metadata = model.metadata()?;
+            if let Some(aivm_style_vectors) = metadata.custom("aivm_style_vectors")? {
+                let aivm_style_vectors = BASE64_STANDARD.decode(aivm_style_vectors)?;
+                let style_vectors = Cursor::new(&aivm_style_vectors);
+                let reader = npyz::NpyFile::new(style_vectors)?;
+                let style_vectors = {
+                    let shape = reader.shape().to_vec();
+                    let order = reader.order();
+                    let data = reader.into_vec::<f32>()?;
+                    let shape = match shape[..] {
+                        [i1, i2] => [i1 as usize, i2 as usize],
+                        _ => panic!("expected 2D array"),
+                    };
+                    let true_shape = shape.set_f(order == npyz::Order::Fortran);
+                    ndarray::Array2::from_shape_vec(true_shape, data)?
+                };
+                drop(metadata);
+                self.models.push(TTSModel {
+                    vits2: if load { Some(model) } else { None },
+                    bytes: if self.max_loaded_models.is_some() {
+                        Some(aivmx_bytes.as_ref().to_vec())
+                    } else {
+                        None
+                    },
+                    ident,
+                    style_vectors,
+                })
+            }
+        }
+        Ok(())
+    }
+
+    /// Load a .sbv2 file binary
+    ///
+    /// # Examples
+    ///
+    /// ```rs
+    /// tts_holder.load_sbv2file("tsukuyomi", std::fs::read("tsukuyomi.sbv2")?)?;
+    /// ```
+    pub fn load_sbv2file<I: Into<TTSIdent>, P: AsRef<[u8]>>(
+        &mut self,
+        ident: I,
+        sbv2_bytes: P,
+    ) -> Result<()> {
+        let (style_vectors, vits2) = crate::sbv2file::parse_sbv2file(sbv2_bytes)?;
+        self.load(ident, style_vectors, vits2)?;
+        Ok(())
+    }
+
+    /// Load a style vector and onnx model binary
+    ///
+    /// # Examples
+    ///
+    /// ```rs
+    /// tts_holder.load("tsukuyomi", std::fs::read("style_vectors.json")?, std::fs::read("model.onnx")?)?;
+    /// ```
+    pub fn load<I: Into<TTSIdent>, P: AsRef<[u8]>>(
+        &mut self,
+        ident: I,
+        style_vectors_bytes: P,
+        vits2_bytes: P,
+    ) -> Result<()> {
+        let ident = ident.into();
+        if self.find_model(ident.clone()).is_err() {
+            let mut load = true;
+            if let Some(max) = self.max_loaded_models {
+                if self.models.iter().filter(|x| x.vits2.is_some()).count() >= max {
+                    load = false;
+                }
+            }
+            self.models.push(TTSModel {
+                vits2: if load {
+                    Some(model::load_model(&vits2_bytes, false)?)
+                } else {
+                    None
+                },
+                style_vectors: style::load_style(style_vectors_bytes)?,
+                ident,
+                bytes: if self.max_loaded_models.is_some() {
+                    Some(vits2_bytes.as_ref().to_vec())
+                } else {
+                    None
+                },
+            })
+        }
+        Ok(())
+    }
+
+    /// Unload a model
+    pub fn unload<I: Into<TTSIdent>>(&mut self, ident: I) -> bool {
+        let ident = ident.into();
+        if let Some((i, _)) = self
+            .models
+            .iter()
+            .enumerate()
+            .find(|(_, m)| m.ident == ident)
+        {
+            self.models.remove(i);
+            true
+        } else {
+            false
+        }
+    }
+
+    /// Parse text and return the input for synthesize
+    ///
+    /// # Note
+    /// This function is for low-level usage, use `easy_synthesize` for high-level usage.
+    #[allow(clippy::type_complexity)]
+    pub fn parse_text(
+        &mut self,
+        text: &str,
+    ) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
+        crate::tts_util::parse_text_blocking(
+            text,
+            None,
+            &self.jtalk,
+            &self.tokenizer,
+            |token_ids, attention_masks| {
+                crate::bert::predict(&mut self.bert, token_ids, attention_masks)
+            },
+        )
+    }
+
+    #[allow(clippy::type_complexity)]
+    pub fn parse_text_neo(
+        &mut self,
+        text: String,
+        given_tones: Option<Vec<i32>>,
+    ) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
+        crate::tts_util::parse_text_blocking(
+            &text,
+            given_tones,
+            &self.jtalk,
+            &self.tokenizer,
+            |token_ids, attention_masks| {
+                crate::bert::predict(&mut self.bert, token_ids, attention_masks)
+            },
+        )
+    }
+
+    fn find_model<I: Into<TTSIdent>>(&mut self, ident: I) -> Result<&mut TTSModel> {
+        let ident = ident.into();
+        self.models
+            .iter_mut()
+            .find(|m| m.ident == ident)
+            .ok_or(Error::ModelNotFoundError(ident.to_string()))
+    }
+    fn find_and_load_model<I: Into<TTSIdent>>(&mut self, ident: I) -> Result<bool> {
+        let ident = ident.into();
+        // Locate target model entry
+        let target_index = self
+            .models
+            .iter()
+            .position(|m| m.ident == ident)
+            .ok_or(Error::ModelNotFoundError(ident.to_string()))?;
+
+        // Already loaded
+        if self.models[target_index].vits2.is_some() {
+            return Ok(true);
+        }
+
+        // Get bytes to build a Session
+        let bytes = self.models[target_index]
+            .bytes
+            .clone()
+            .ok_or(Error::ModelNotFoundError(ident.to_string()))?;
+
+        // Enforce max loaded models by evicting a different loaded model's session, not removing the entry
+        if let Some(max) = self.max_loaded_models {
+            let loaded_count = self.models.iter().filter(|m| m.vits2.is_some()).count();
+            if loaded_count >= max {
+                if let Some(evict_index) = self
+                    .models
+                    .iter()
+                    .position(|m| m.vits2.is_some() && m.ident != ident)
+                {
+                    // Drop only the session to free memory; keep bytes/style for future reload
+                    self.models[evict_index].vits2 = None;
+                }
+            }
+        }
+
+        // Build and set session in-place for the target model
+        let s = model::load_model(&bytes, false)?;
+        self.models[target_index].vits2 = Some(s);
+        Ok(true)
+    }
+
+    /// Get style vector by style id and weight
+    ///
+    /// # Note
+    /// This function is for low-level usage, use `easy_synthesize` for high-level usage.
+    pub fn get_style_vector<I: Into<TTSIdent>>(
+        &mut self,
+        ident: I,
+        style_id: i32,
+        weight: f32,
+    ) -> Result<Array1<f32>> {
+        style::get_style_vector(&self.find_model(ident)?.style_vectors, style_id, weight)
+    }
+
+    /// Synthesize text to audio
+    ///
+    /// # Examples
+    ///
+    /// ```rs
+    /// let audio = tts_holder.easy_synthesize("tsukuyomi", "こんにちは", 0, SynthesizeOptions::default())?;
+    /// ```
+    pub fn easy_synthesize<I: Into<TTSIdent> + Copy>(
+        &mut self,
+        ident: I,
+        text: &str,
+        style_id: i32,
+        speaker_id: i64,
+        options: SynthesizeOptions,
+    ) -> Result<Vec<u8>> {
+        self.find_and_load_model(ident)?;
+        let style_vector = self.get_style_vector(ident, style_id, options.style_weight)?;
+        let audio_array = if options.split_sentences {
+            let texts: Vec<&str> = text.split('\n').collect();
+            let mut audios = vec![];
+            for (i, t) in texts.iter().enumerate() {
+                if t.is_empty() {
+                    continue;
+                }
+                let (bert_ori, phones, tones, lang_ids) = self.parse_text(t)?;
+
+                let vits2 = self
+                    .find_model(ident)?
+                    .vits2
+                    .as_mut()
+                    .ok_or(Error::ModelNotFoundError(ident.into().to_string()))?;
+                let audio = model::synthesize(
+                    vits2,
+                    bert_ori.to_owned(),
+                    phones,
+                    Array1::from_vec(vec![speaker_id]),
+                    tones,
+                    lang_ids,
+                    style_vector.clone(),
+                    options.sdp_ratio,
+                    options.length_scale,
+                    0.677,
+                    0.8,
+                )?;
+                audios.push(audio.clone());
+                if i != texts.len() - 1 {
+                    audios.push(Array3::zeros((1, 1, 22050)));
+                }
+            }
+            concatenate(
+                Axis(2),
+                &audios.iter().map(|x| x.view()).collect::<Vec<_>>(),
+            )?
+        } else {
+            let (bert_ori, phones, tones, lang_ids) = self.parse_text(text)?;
+
+            let vits2 = self
+                .find_model(ident)?
+                .vits2
+                .as_mut()
+                .ok_or(Error::ModelNotFoundError(ident.into().to_string()))?;
+            model::synthesize(
+                vits2,
+                bert_ori.to_owned(),
+                phones,
+                Array1::from_vec(vec![speaker_id]),
+                tones,
+                lang_ids,
+                style_vector,
+                options.sdp_ratio,
+                options.length_scale,
+                0.677,
+                0.8,
+            )?
+        };
+        tts_util::array_to_vec(audio_array)
+    }
+
+    pub fn easy_synthesize_neo<I: Into<TTSIdent> + Copy>(
+        &mut self,
+        ident: I,
+        text: &str,
+        given_tones: Option<Vec<i32>>,
+        style_id: i32,
+        speaker_id: i64,
+        options: SynthesizeOptions,
+    ) -> Result<Vec<u8>> {
+        self.find_and_load_model(ident)?;
+        let style_vector = self.get_style_vector(ident, style_id, options.style_weight)?;
+        let audio_array = if options.split_sentences {
+            let texts: Vec<&str> = text.split('\n').collect();
+            let mut audios = vec![];
+            for (i, t) in texts.iter().enumerate() {
+                if t.is_empty() {
+                    continue;
+                }
+                let (bert_ori, phones, tones, lang_ids) =
+                    self.parse_text_neo(t.to_string(), given_tones.clone())?;
+
+                let vits2 = self
+                    .find_model(ident)?
+                    .vits2
+                    .as_mut()
+                    .ok_or(Error::ModelNotFoundError(ident.into().to_string()))?;
+                let audio = model::synthesize(
+                    vits2,
+                    bert_ori.to_owned(),
+                    phones,
+                    Array1::from_vec(vec![speaker_id]),
+                    tones,
+                    lang_ids,
+                    style_vector.clone(),
+                    options.sdp_ratio,
+                    options.length_scale,
+                    0.677,
+                    0.8,
+                )?;
+                audios.push(audio.clone());
+                if i != texts.len() - 1 {
+                    audios.push(Array3::zeros((1, 1, 22050)));
+                }
+            }
+            concatenate(
+                Axis(2),
+                &audios.iter().map(|x| x.view()).collect::<Vec<_>>(),
+            )?
+        } else {
+            let (bert_ori, phones, tones, lang_ids) = self.parse_text(text)?;
+
+            let vits2 = self
+                .find_model(ident)?
+                .vits2
+                .as_mut()
+                .ok_or(Error::ModelNotFoundError(ident.into().to_string()))?;
+            model::synthesize(
+                vits2,
+                bert_ori.to_owned(),
+                phones,
+                Array1::from_vec(vec![speaker_id]),
+                tones,
+                lang_ids,
+                style_vector,
+                options.sdp_ratio,
+                options.length_scale,
+                0.677,
+                0.8,
+            )?
+        };
+        tts_util::array_to_vec(audio_array)
+    }
+}
+
+/// Synthesize options
+///
+/// # Fields
+/// - `sdp_ratio`: SDP ratio
+/// - `length_scale`: Length scale
+/// - `style_weight`: Style weight
+/// - `split_sentences`: Split sentences
+pub struct SynthesizeOptions {
+    pub sdp_ratio: f32,
+    pub length_scale: f32,
+    pub style_weight: f32,
+    pub split_sentences: bool,
+}
+
+impl Default for SynthesizeOptions {
+    fn default() -> Self {
+        SynthesizeOptions {
+            sdp_ratio: 0.0,
+            length_scale: 1.0,
+            style_weight: 1.0,
+            split_sentences: true,
+        }
+    }
+}

crates/sbv2_core/src/tts_util.rs

@@ -0,0 +1,227 @@
+use std::io::Cursor;
+
+use crate::error::Result;
+use crate::jtalk::JTalkProcess;
+use crate::mora::MORA_KATA_TO_MORA_PHONEMES;
+use crate::norm::PUNCTUATIONS;
+use crate::{jtalk, nlp, norm, tokenizer, utils};
+use hound::{SampleFormat, WavSpec, WavWriter};
+use ndarray::{concatenate, s, Array, Array1, Array2, Array3, Axis};
+use tokenizers::Tokenizer;
+
+pub fn preprocess_parse_text(text: &str, jtalk: &jtalk::JTalk) -> Result<(String, JTalkProcess)> {
+    let text = jtalk.num2word(text)?;
+    let normalized_text = norm::normalize_text(&text);
+
+    let process = jtalk.process_text(&normalized_text)?;
+    Ok((normalized_text, process))
+}
+
+/// Parse text and return the input for synthesize
+///
+/// # Note
+/// This function is for low-level usage, use `easy_synthesize` for high-level usage.
+#[allow(clippy::type_complexity)]
+pub async fn parse_text(
+    text: &str,
+    jtalk: &jtalk::JTalk,
+    tokenizer: &Tokenizer,
+    bert_predict: impl FnOnce(
+        Vec<i64>,
+        Vec<i64>,
+    ) -> std::pin::Pin<
+        Box<dyn std::future::Future<Output = Result<ndarray::Array2<f32>>>>,
+    >,
+) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
+    let (normalized_text, process) = preprocess_parse_text(text, jtalk)?;
+    let (phones, tones, mut word2ph) = process.g2p()?;
+    let (phones, tones, lang_ids) = nlp::cleaned_text_to_sequence(phones, tones);
+    let phones = utils::intersperse(&phones, 0);
+    let tones = utils::intersperse(&tones, 0);
+    let lang_ids = utils::intersperse(&lang_ids, 0);
+    for item in &mut word2ph {
+        *item *= 2;
+    }
+    word2ph[0] += 1;
+
+    let text = {
+        let (seq_text, _) = process.text_to_seq_kata()?;
+        seq_text.join("")
+    };
+    let (token_ids, attention_masks) = tokenizer::tokenize(&text, tokenizer)?;
+
+    let bert_content = bert_predict(token_ids, attention_masks).await?;
+
+    assert!(
+        word2ph.len() == text.chars().count() + 2,
+        "{} {}",
+        word2ph.len(),
+        normalized_text.chars().count()
+    );
+
+    let mut phone_level_feature = vec![];
+    for (i, reps) in word2ph.iter().enumerate() {
+        let repeat_feature = {
+            let (reps_rows, reps_cols) = (*reps, 1);
+            let arr_len = bert_content.slice(s![i, ..]).len();
+
+            let mut results: Array2<f32> = Array::zeros((reps_rows as usize, arr_len * reps_cols));
+
+            for j in 0..reps_rows {
+                for k in 0..reps_cols {
+                    let mut view = results.slice_mut(s![j, k * arr_len..(k + 1) * arr_len]);
+                    view.assign(&bert_content.slice(s![i, ..]));
+                }
+            }
+            results
+        };
+        phone_level_feature.push(repeat_feature);
+    }
+    let phone_level_feature = concatenate(
+        Axis(0),
+        &phone_level_feature
+            .iter()
+            .map(|x| x.view())
+            .collect::<Vec<_>>(),
+    )?;
+    let bert_ori = phone_level_feature.t();
+    Ok((
+        bert_ori.to_owned(),
+        phones.into(),
+        tones.into(),
+        lang_ids.into(),
+    ))
+}
+
+/// Parse text and return the input for synthesize
+///
+/// # Note
+/// This function is for low-level usage, use `easy_synthesize` for high-level usage.
+#[allow(clippy::type_complexity)]
+pub fn parse_text_blocking(
+    text: &str,
+    given_tones: Option<Vec<i32>>,
+    jtalk: &jtalk::JTalk,
+    tokenizer: &Tokenizer,
+    bert_predict: impl FnOnce(Vec<i64>, Vec<i64>) -> Result<ndarray::Array2<f32>>,
+) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
+    let text = jtalk.num2word(text)?;
+    let normalized_text = norm::normalize_text(&text);
+
+    let process = jtalk.process_text(&normalized_text)?;
+    let (phones, mut tones, mut word2ph) = process.g2p()?;
+    if let Some(given_tones) = given_tones {
+        tones = given_tones;
+    }
+    let (phones, tones, lang_ids) = nlp::cleaned_text_to_sequence(phones, tones);
+
+    let phones = utils::intersperse(&phones, 0);
+    let tones = utils::intersperse(&tones, 0);
+    let lang_ids = utils::intersperse(&lang_ids, 0);
+    for item in &mut word2ph {
+        *item *= 2;
+    }
+    word2ph[0] += 1;
+
+    let text = {
+        let (seq_text, _) = process.text_to_seq_kata()?;
+        seq_text.join("")
+    };
+    let (token_ids, attention_masks) = tokenizer::tokenize(&text, tokenizer)?;
+
+    let bert_content = bert_predict(token_ids, attention_masks)?;
+
+    assert!(
+        word2ph.len() == text.chars().count() + 2,
+        "{} {}",
+        word2ph.len(),
+        normalized_text.chars().count()
+    );
+
+    let mut phone_level_feature = vec![];
+    for (i, reps) in word2ph.iter().enumerate() {
+        let repeat_feature = {
+            let (reps_rows, reps_cols) = (*reps, 1);
+            let arr_len = bert_content.slice(s![i, ..]).len();
+
+            let mut results: Array2<f32> = Array::zeros((reps_rows as usize, arr_len * reps_cols));
+
+            for j in 0..reps_rows {
+                for k in 0..reps_cols {
+                    let mut view = results.slice_mut(s![j, k * arr_len..(k + 1) * arr_len]);
+                    view.assign(&bert_content.slice(s![i, ..]));
+                }
+            }
+            results
+        };
+        phone_level_feature.push(repeat_feature);
+    }
+    let phone_level_feature = concatenate(
+        Axis(0),
+        &phone_level_feature
+            .iter()
+            .map(|x| x.view())
+            .collect::<Vec<_>>(),
+    )?;
+    let bert_ori = phone_level_feature.t();
+    Ok((
+        bert_ori.to_owned(),
+        phones.into(),
+        tones.into(),
+        lang_ids.into(),
+    ))
+}
+
+pub fn array_to_vec(audio_array: Array3<f32>) -> Result<Vec<u8>> {
+    // If SBV2_FORCE_STEREO is set ("1"/"true"), duplicate mono to stereo
+    let force_stereo = std::env::var("SBV2_FORCE_STEREO")
+        .ok()
+        .map(|v| matches!(v.as_str(), "1" | "true" | "TRUE" | "True"))
+        .unwrap_or(false);
+
+    let channels: u16 = if force_stereo { 2 } else { 1 };
+    let spec = WavSpec {
+        channels,
+        sample_rate: 44100,
+        bits_per_sample: 32,
+        sample_format: SampleFormat::Float,
+    };
+    let mut cursor = Cursor::new(Vec::new());
+    let mut writer = WavWriter::new(&mut cursor, spec)?;
+    for i in 0..audio_array.shape()[0] {
+        let output = audio_array.slice(s![i, 0, ..]).to_vec();
+        if force_stereo {
+            for sample in output {
+                // Write to Left and Right channels
+                writer.write_sample(sample)?;
+                writer.write_sample(sample)?;
+            }
+        } else {
+            for sample in output {
+                writer.write_sample(sample)?;
+            }
+        }
+    }
+    writer.finalize()?;
+    Ok(cursor.into_inner())
+}
+
+pub fn kata_tone2phone_tone(kata_tone: Vec<(String, i32)>) -> Vec<(String, i32)> {
+    let mut results = vec![("_".to_string(), 0)];
+    for (mora, tone) in kata_tone {
+        if PUNCTUATIONS.contains(&mora.as_str()) {
+            results.push((mora, 0));
+            continue;
+        } else {
+            let (consonant, vowel) = MORA_KATA_TO_MORA_PHONEMES.get(&mora).unwrap();
+            if let Some(consonant) = consonant {
+                results.push((consonant.to_string(), tone));
+                results.push((vowel.to_string(), tone));
+            } else {
+                results.push((vowel.to_string(), tone));
+            }
+        }
+    }
+    results.push(("_".to_string(), 0));
+    results
+}

crates/sbv2_editor/Cargo.toml

@@ -0,0 +1,19 @@
+[package]
+name = "sbv2_editor"
+version.workspace = true
+edition.workspace = true
+description.workspace = true
+license.workspace = true
+readme.workspace = true
+repository.workspace = true
+documentation.workspace = true
+
+[dependencies]
+anyhow.workspace = true
+axum = "0.8.1"
+dotenvy.workspace = true
+env_logger.workspace = true
+log = "0.4.27"
+sbv2_core = { version = "0.2.0-alpha6", path = "../sbv2_core", features = ["aivmx"] }
+serde = { version = "1.0.219", features = ["derive"] }
+tokio = { version = "1.44.1", features = ["full"] }

crates/sbv2_editor/README.md

@@ -0,0 +1,2 @@
+# sbv2-voicevox
+sbv2-apiをvoicevox化します。

crates/sbv2_editor/query2.json

@@ -0,0 +1,226 @@
+{
+    "accent_phrases": [
+        {
+            "moras": [
+                {
+                    "text": "コ",
+                    "consonant": "k",
+                    "consonant_length": 0.10002632439136505,
+                    "vowel": "o",
+                    "vowel_length": 0.15740256011486053,
+                    "pitch": 5.749961853027344
+                },
+                {
+                    "text": "ン",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "N",
+                    "vowel_length": 0.08265873789787292,
+                    "pitch": 5.89122200012207
+                },
+                {
+                    "text": "ニ",
+                    "consonant": "n",
+                    "consonant_length": 0.03657080978155136,
+                    "vowel": "i",
+                    "vowel_length": 0.1175866425037384,
+                    "pitch": 5.969866752624512
+                },
+                {
+                    "text": "チ",
+                    "consonant": "ch",
+                    "consonant_length": 0.09005842357873917,
+                    "vowel": "i",
+                    "vowel_length": 0.08666137605905533,
+                    "pitch": 5.958892822265625
+                },
+                {
+                    "text": "ワ",
+                    "consonant": "w",
+                    "consonant_length": 0.07833231985569,
+                    "vowel": "a",
+                    "vowel_length": 0.21250136196613312,
+                    "pitch": 5.949411392211914
+                }
+            ],
+            "accent": 5,
+            "pause_mora": {
+                "text": "、",
+                "consonant": null,
+                "consonant_length": null,
+                "vowel": "pau",
+                "vowel_length": 0.4723339378833771,
+                "pitch": 0.0
+            },
+            "is_interrogative": false
+        },
+        {
+            "moras": [
+                {
+                    "text": "オ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "o",
+                    "vowel_length": 0.22004225850105286,
+                    "pitch": 5.6870927810668945
+                },
+                {
+                    "text": "ン",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "N",
+                    "vowel_length": 0.09161105751991272,
+                    "pitch": 5.93472957611084
+                },
+                {
+                    "text": "セ",
+                    "consonant": "s",
+                    "consonant_length": 0.08924821764230728,
+                    "vowel": "e",
+                    "vowel_length": 0.14142127335071564,
+                    "pitch": 6.121850490570068
+                },
+                {
+                    "text": "エ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "e",
+                    "vowel_length": 0.10636933892965317,
+                    "pitch": 6.157896041870117
+                },
+                {
+                    "text": "ゴ",
+                    "consonant": "g",
+                    "consonant_length": 0.07600915431976318,
+                    "vowel": "o",
+                    "vowel_length": 0.09598273783922195,
+                    "pitch": 6.188933849334717
+                },
+                {
+                    "text": "オ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "o",
+                    "vowel_length": 0.1079121008515358,
+                    "pitch": 6.235202789306641
+                },
+                {
+                    "text": "セ",
+                    "consonant": "s",
+                    "consonant_length": 0.09591838717460632,
+                    "vowel": "e",
+                    "vowel_length": 0.10286372154951096,
+                    "pitch": 6.153214454650879
+                },
+                {
+                    "text": "エ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "e",
+                    "vowel_length": 0.08992656320333481,
+                    "pitch": 6.02571439743042
+                },
+                {
+                    "text": "ノ",
+                    "consonant": "n",
+                    "consonant_length": 0.05660202354192734,
+                    "vowel": "o",
+                    "vowel_length": 0.09676017612218857,
+                    "pitch": 5.711844444274902
+                }
+            ],
+            "accent": 5,
+            "pause_mora": null,
+            "is_interrogative": false
+        },
+        {
+            "moras": [
+                {
+                    "text": "セ",
+                    "consonant": "s",
+                    "consonant_length": 0.07805486768484116,
+                    "vowel": "e",
+                    "vowel_length": 0.09617523103952408,
+                    "pitch": 5.774399280548096
+                },
+                {
+                    "text": "カ",
+                    "consonant": "k",
+                    "consonant_length": 0.06712044775485992,
+                    "vowel": "a",
+                    "vowel_length": 0.148829385638237,
+                    "pitch": 6.063965797424316
+                },
+                {
+                    "text": "イ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "i",
+                    "vowel_length": 0.11061104387044907,
+                    "pitch": 6.040698051452637
+                },
+                {
+                    "text": "エ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "e",
+                    "vowel_length": 0.13046696782112122,
+                    "pitch": 5.806027889251709
+                }
+            ],
+            "accent": 1,
+            "pause_mora": null,
+            "is_interrogative": false
+        },
+        {
+            "moras": [
+                {
+                    "text": "ヨ",
+                    "consonant": "y",
+                    "consonant_length": 0.07194744795560837,
+                    "vowel": "o",
+                    "vowel_length": 0.08622600883245468,
+                    "pitch": 5.694094657897949
+                },
+                {
+                    "text": "オ",
+                    "consonant": null,
+                    "consonant_length": null,
+                    "vowel": "o",
+                    "vowel_length": 0.10635452717542648,
+                    "pitch": 5.787222385406494
+                },
+                {
+                    "text": "コ",
+                    "consonant": "k",
+                    "consonant_length": 0.07077334076166153,
+                    "vowel": "o",
+                    "vowel_length": 0.09248624742031097,
+                    "pitch": 5.793357849121094
+                },
+                {
+                    "text": "ソ",
+                    "consonant": "s",
+                    "consonant_length": 0.08705667406320572,
+                    "vowel": "o",
+                    "vowel_length": 0.2238258570432663,
+                    "pitch": 5.643765449523926
+                }
+            ],
+            "accent": 1,
+            "pause_mora": null,
+            "is_interrogative": false
+        }
+    ],
+    "speedScale": 1.0,
+    "pitchScale": 0.0,
+    "intonationScale": 1.0,
+    "volumeScale": 1.0,
+    "prePhonemeLength": 0.1,
+    "postPhonemeLength": 0.1,
+    "pauseLength": null,
+    "pauseLengthScale": 1.0,
+    "outputSamplingRate": 24000,
+    "outputStereo": false,
+    "kana": "コンニチワ'、オンセエゴ'オセエノ/セ'カイエ/ヨ'オコソ"
+}

crates/sbv2_editor/src/error.rs

@@ -0,0 +1,27 @@
+use axum::{
+    http::StatusCode,
+    response::{IntoResponse, Response},
+};
+
+pub type AppResult<T> = std::result::Result<T, AppError>;
+
+pub struct AppError(anyhow::Error);
+
+impl IntoResponse for AppError {
+    fn into_response(self) -> Response {
+        (
+            StatusCode::INTERNAL_SERVER_ERROR,
+            format!("Something went wrong: {}", self.0),
+        )
+            .into_response()
+    }
+}
+
+impl<E> From<E> for AppError
+where
+    E: Into<anyhow::Error>,
+{
+    fn from(err: E) -> Self {
+        Self(err.into())
+    }
+}

crates/sbv2_editor/src/main.rs

@@ -0,0 +1,197 @@
+use axum::extract::State;
+use axum::{
+    extract::Query,
+    http::header::CONTENT_TYPE,
+    response::IntoResponse,
+    routing::{get, post},
+    Json, Router,
+};
+use sbv2_core::tts_util::kata_tone2phone_tone;
+use sbv2_core::{
+    tts::{SynthesizeOptions, TTSModelHolder},
+    tts_util::preprocess_parse_text,
+};
+use serde::{Deserialize, Serialize};
+use tokio::{fs, net::TcpListener, sync::Mutex};
+
+use std::env;
+use std::sync::Arc;
+
+use error::AppResult;
+
+mod error;
+
+#[derive(Deserialize)]
+struct RequestCreateAudioQuery {
+    text: String,
+}
+
+#[derive(Serialize, Deserialize)]
+struct AudioQuery {
+    kana: String,
+    tone: i32,
+}
+
+#[derive(Serialize)]
+struct ResponseCreateAudioQuery {
+    audio_query: Vec<AudioQuery>,
+    text: String,
+}
+
+async fn create_audio_query(
+    State(state): State<AppState>,
+    Query(request): Query<RequestCreateAudioQuery>,
+) -> AppResult<impl IntoResponse> {
+    let (text, process) = {
+        let tts_model = state.tts_model.lock().await;
+        preprocess_parse_text(&request.text, &tts_model.jtalk)?
+    };
+    let kana_tone_list = process.g2kana_tone()?;
+    let audio_query = kana_tone_list
+        .iter()
+        .map(|(kana, tone)| AudioQuery {
+            kana: kana.clone(),
+            tone: *tone,
+        })
+        .collect::<Vec<_>>();
+    Ok(Json(ResponseCreateAudioQuery { audio_query, text }))
+}
+
+#[derive(Deserialize)]
+pub struct RequestSynthesis {
+    text: String,
+    speaker_id: i64,
+    sdp_ratio: f32,
+    length_scale: f32,
+    style_id: i32,
+    audio_query: Vec<AudioQuery>,
+    ident: String,
+}
+
+async fn synthesis(
+    State(state): State<AppState>,
+    Json(request): Json<RequestSynthesis>,
+) -> AppResult<impl IntoResponse> {
+    let phone_tone = request
+        .audio_query
+        .iter()
+        .map(|query| (query.kana.clone(), query.tone))
+        .collect::<Vec<_>>();
+    let phone_tone = kata_tone2phone_tone(phone_tone);
+    let tones = phone_tone.iter().map(|(_, tone)| *tone).collect::<Vec<_>>();
+    let buffer = {
+        let mut tts_model = state.tts_model.lock().await;
+        tts_model.easy_synthesize_neo(
+            &request.ident,
+            &request.text,
+            Some(tones),
+            request.style_id,
+            request.speaker_id,
+            SynthesizeOptions {
+                sdp_ratio: request.sdp_ratio,
+                length_scale: request.length_scale,
+                ..Default::default()
+            },
+        )?
+    };
+    Ok(([(CONTENT_TYPE, "audio/wav")], buffer))
+}
+
+#[derive(Clone)]
+struct AppState {
+    tts_model: Arc<Mutex<TTSModelHolder>>,
+}
+
+impl AppState {
+    pub async fn new() -> anyhow::Result<Self> {
+        let mut tts_model = TTSModelHolder::new(
+            &fs::read(env::var("BERT_MODEL_PATH")?).await?,
+            &fs::read(env::var("TOKENIZER_PATH")?).await?,
+            env::var("HOLDER_MAX_LOADED_MODElS")
+                .ok()
+                .and_then(|x| x.parse().ok()),
+        )?;
+        let models = env::var("MODELS_PATH").unwrap_or("models".to_string());
+        let mut f = fs::read_dir(&models).await?;
+        let mut entries = vec![];
+        while let Ok(Some(e)) = f.next_entry().await {
+            let name = e.file_name().to_string_lossy().to_string();
+            if name.ends_with(".onnx") && name.starts_with("model_") {
+                let name_len = name.len();
+                let name = name.chars();
+                entries.push(
+                    name.collect::<Vec<_>>()[6..name_len - 5]
+                        .iter()
+                        .collect::<String>(),
+                );
+            } else if name.ends_with(".sbv2") {
+                let entry = &name[..name.len() - 5];
+                log::info!("Try loading: {entry}");
+                let sbv2_bytes = match fs::read(format!("{models}/{entry}.sbv2")).await {
+                    Ok(b) => b,
+                    Err(e) => {
+                        log::warn!("Error loading sbv2_bytes from file {entry}: {e}");
+                        continue;
+                    }
+                };
+                if let Err(e) = tts_model.load_sbv2file(entry, sbv2_bytes) {
+                    log::warn!("Error loading {entry}: {e}");
+                };
+                log::info!("Loaded: {entry}");
+            } else if name.ends_with(".aivmx") {
+                let entry = &name[..name.len() - 6];
+                log::info!("Try loading: {entry}");
+                let aivmx_bytes = match fs::read(format!("{models}/{entry}.aivmx")).await {
+                    Ok(b) => b,
+                    Err(e) => {
+                        log::warn!("Error loading aivmx bytes from file {entry}: {e}");
+                        continue;
+                    }
+                };
+                if let Err(e) = tts_model.load_aivmx(entry, aivmx_bytes) {
+                    log::error!("Error loading {entry}: {e}");
+                }
+                log::info!("Loaded: {entry}");
+            }
+        }
+        for entry in entries {
+            log::info!("Try loading: {entry}");
+            let style_vectors_bytes =
+                match fs::read(format!("{models}/style_vectors_{entry}.json")).await {
+                    Ok(b) => b,
+                    Err(e) => {
+                        log::warn!("Error loading style_vectors_bytes from file {entry}: {e}");
+                        continue;
+                    }
+                };
+            let vits2_bytes = match fs::read(format!("{models}/model_{entry}.onnx")).await {
+                Ok(b) => b,
+                Err(e) => {
+                    log::warn!("Error loading vits2_bytes from file {entry}: {e}");
+                    continue;
+                }
+            };
+            if let Err(e) = tts_model.load(&entry, style_vectors_bytes, vits2_bytes) {
+                log::warn!("Error loading {entry}: {e}");
+            };
+            log::info!("Loaded: {entry}");
+        }
+        Ok(Self {
+            tts_model: Arc::new(Mutex::new(tts_model)),
+        })
+    }
+}
+
+#[tokio::main]
+async fn main() -> anyhow::Result<()> {
+    dotenvy::dotenv_override().ok();
+    env_logger::init();
+    let app = Router::new()
+        .route("/", get(|| async { "Hello, world!" }))
+        .route("/audio_query", get(create_audio_query))
+        .route("/synthesis", post(synthesis))
+        .with_state(AppState::new().await?);
+    let listener = TcpListener::bind("0.0.0.0:8080").await?;
+    axum::serve(listener, app).await?;
+    Ok(())
+}

crates/sbv2_wasm/Cargo.toml

@@ -0,0 +1,20 @@
+[package]
+name = "sbv2_wasm"
+version.workspace = true
+edition.workspace = true
+description.workspace = true
+readme.workspace = true
+repository.workspace = true
+documentation.workspace = true
+license.workspace = true
+
+[lib]
+crate-type = ["cdylib", "rlib"]
+
+[dependencies]
+wasm-bindgen = "0.2.93"
+sbv2_core = { path = "../sbv2_core", default-features = false, features = ["no_std"] }
+once_cell.workspace = true
+js-sys = "0.3.70"
+ndarray.workspace = true
+wasm-bindgen-futures = "0.4.43"

crates/sbv2_wasm/README.md

@@ -0,0 +1,2 @@
+# StyleBertVITS2 wasm
+refer to https://github.com/neodyland/sbv2-api

crates/sbv2_wasm/biome.json

@@ -0,0 +1,31 @@
+{
+	"$schema": "https://biomejs.dev/schemas/1.9.2/schema.json",
+	"vcs": {
+		"enabled": false,
+		"clientKind": "git",
+		"useIgnoreFile": false
+	},
+	"files": {
+		"ignoreUnknown": false,
+		"ignore": []
+	},
+	"formatter": {
+		"enabled": true,
+		"indentStyle": "tab",
+		"ignore": ["dist/", "pkg/"]
+	},
+	"organizeImports": {
+		"enabled": true
+	},
+	"linter": {
+		"enabled": true,
+		"rules": {
+			"recommended": true
+		}
+	},
+	"javascript": {
+		"formatter": {
+			"quoteStyle": "double"
+		}
+	}
+}

crates/sbv2_wasm/build.sh

@@ -0,0 +1,8 @@
+#!/bin/sh
+wasm-pack build --target web ./crates/sbv2_wasm --release
+wasm-opt -O3 -o ./crates/sbv2_wasm/pkg/sbv2_wasm_bg.wasm ./crates/sbv2_wasm/pkg/sbv2_wasm_bg.wasm
+wasm-strip ./crates/sbv2_wasm/pkg/sbv2_wasm_bg.wasm
+mkdir -p ./crates/sbv2_wasm/dist
+cp ./crates/sbv2_wasm/pkg/sbv2_wasm_bg.wasm ./crates/sbv2_wasm/dist/sbv2_wasm_bg.wasm
+cd ./crates/sbv2_wasm
+pnpm build

crates/sbv2_wasm/example.js

@@ -0,0 +1,14 @@
+import { ModelHolder } from "./dist/index.js";
+import fs from "node:fs/promises";
+
+ModelHolder.globalInit(await fs.readFile("./dist/sbv2_wasm_bg.wasm"));
+const holder = await ModelHolder.create(
+	(await fs.readFile("../../models/tokenizer.json")).toString("utf-8"),
+	await fs.readFile("../../models/deberta.onnx"),
+);
+await holder.load(
+	"tsukuyomi",
+	await fs.readFile("../../models/tsukuyomi.sbv2"),
+);
+await fs.writeFile("out.wav", await holder.synthesize("tsukuyomi", "おはよう"));
+holder.unload("tsukuyomi");

crates/sbv2_wasm/package.json

@@ -0,0 +1,26 @@
+{
+	"name": "sbv2",
+	"version": "0.2.0-alpha6",
+	"description": "Style Bert VITS2 wasm",
+	"main": "dist/index.js",
+	"types": "dist/index.d.ts",
+	"type": "module",
+	"scripts": {
+		"build": "tsc && esbuild src-js/index.ts --outfile=dist/index.js --minify --format=esm --bundle --external:onnxruntime-web",
+		"format": "biome format --write ."
+	},
+	"keywords": [],
+	"author": "tuna2134",
+        "contributes": ["neodyland"],
+	"license": "MIT",
+	"devDependencies": {
+		"@biomejs/biome": "^1.9.4",
+		"@types/node": "^22.13.5",
+		"esbuild": "^0.25.0",
+		"typescript": "^5.7.3"
+	},
+	"dependencies": {
+		"onnxruntime-web": "^1.20.1"
+	},
+	"files": ["dist/*", "package.json", "README.md", "pkg/*.ts", "pkg/*.js"]
+}

crates/sbv2_wasm/pnpm-lock.yaml

@@ -0,0 +1,504 @@
+lockfileVersion: '9.0'
+
+settings:
+  autoInstallPeers: true
+  excludeLinksFromLockfile: false
+
+importers:
+
+  .:
+    dependencies:
+      onnxruntime-web:
+        specifier: ^1.20.1
+        version: 1.20.1
+    devDependencies:
+      '@biomejs/biome':
+        specifier: ^1.9.4
+        version: 1.9.4
+      '@types/node':
+        specifier: ^22.13.5
+        version: 22.13.5
+      esbuild:
+        specifier: ^0.25.0
+        version: 0.25.0
+      typescript:
+        specifier: ^5.7.3
+        version: 5.8.3
+
+packages:
+
+  '@biomejs/biome@1.9.4':
+    resolution: {integrity: sha512-1rkd7G70+o9KkTn5KLmDYXihGoTaIGO9PIIN2ZB7UJxFrWw04CZHPYiMRjYsaDvVV7hP1dYNRLxSANLaBFGpog==}
+    engines: {node: '>=14.21.3'}
+    hasBin: true
+
+  '@biomejs/cli-darwin-arm64@1.9.4':
+    resolution: {integrity: sha512-bFBsPWrNvkdKrNCYeAp+xo2HecOGPAy9WyNyB/jKnnedgzl4W4Hb9ZMzYNbf8dMCGmUdSavlYHiR01QaYR58cw==}
+    engines: {node: '>=14.21.3'}
+    cpu: [arm64]
+    os: [darwin]
+
+  '@biomejs/cli-darwin-x64@1.9.4':
+    resolution: {integrity: sha512-ngYBh/+bEedqkSevPVhLP4QfVPCpb+4BBe2p7Xs32dBgs7rh9nY2AIYUL6BgLw1JVXV8GlpKmb/hNiuIxfPfZg==}
+    engines: {node: '>=14.21.3'}
+    cpu: [x64]
+    os: [darwin]
+
+  '@biomejs/cli-linux-arm64-musl@1.9.4':
+    resolution: {integrity: sha512-v665Ct9WCRjGa8+kTr0CzApU0+XXtRgwmzIf1SeKSGAv+2scAlW6JR5PMFo6FzqqZ64Po79cKODKf3/AAmECqA==}
+    engines: {node: '>=14.21.3'}
+    cpu: [arm64]
+    os: [linux]
+
+  '@biomejs/cli-linux-arm64@1.9.4':
+    resolution: {integrity: sha512-fJIW0+LYujdjUgJJuwesP4EjIBl/N/TcOX3IvIHJQNsAqvV2CHIogsmA94BPG6jZATS4Hi+xv4SkBBQSt1N4/g==}
+    engines: {node: '>=14.21.3'}
+    cpu: [arm64]
+    os: [linux]
+
+  '@biomejs/cli-linux-x64-musl@1.9.4':
+    resolution: {integrity: sha512-gEhi/jSBhZ2m6wjV530Yy8+fNqG8PAinM3oV7CyO+6c3CEh16Eizm21uHVsyVBEB6RIM8JHIl6AGYCv6Q6Q9Tg==}
+    engines: {node: '>=14.21.3'}
+    cpu: [x64]
+    os: [linux]
+
+  '@biomejs/cli-linux-x64@1.9.4':
+    resolution: {integrity: sha512-lRCJv/Vi3Vlwmbd6K+oQ0KhLHMAysN8lXoCI7XeHlxaajk06u7G+UsFSO01NAs5iYuWKmVZjmiOzJ0OJmGsMwg==}
+    engines: {node: '>=14.21.3'}
+    cpu: [x64]
+    os: [linux]
+
+  '@biomejs/cli-win32-arm64@1.9.4':
+    resolution: {integrity: sha512-tlbhLk+WXZmgwoIKwHIHEBZUwxml7bRJgk0X2sPyNR3S93cdRq6XulAZRQJ17FYGGzWne0fgrXBKpl7l4M87Hg==}
+    engines: {node: '>=14.21.3'}
+    cpu: [arm64]
+    os: [win32]
+
+  '@biomejs/cli-win32-x64@1.9.4':
+    resolution: {integrity: sha512-8Y5wMhVIPaWe6jw2H+KlEm4wP/f7EW3810ZLmDlrEEy5KvBsb9ECEfu/kMWD484ijfQ8+nIi0giMgu9g1UAuuA==}
+    engines: {node: '>=14.21.3'}
+    cpu: [x64]
+    os: [win32]
+
+  '@esbuild/aix-ppc64@0.25.0':
+    resolution: {integrity: sha512-O7vun9Sf8DFjH2UtqK8Ku3LkquL9SZL8OLY1T5NZkA34+wG3OQF7cl4Ql8vdNzM6fzBbYfLaiRLIOZ+2FOCgBQ==}
+    engines: {node: '>=18'}
+    cpu: [ppc64]
+    os: [aix]
+
+  '@esbuild/android-arm64@0.25.0':
+    resolution: {integrity: sha512-grvv8WncGjDSyUBjN9yHXNt+cq0snxXbDxy5pJtzMKGmmpPxeAmAhWxXI+01lU5rwZomDgD3kJwulEnhTRUd6g==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [android]
+
+  '@esbuild/android-arm@0.25.0':
+    resolution: {integrity: sha512-PTyWCYYiU0+1eJKmw21lWtC+d08JDZPQ5g+kFyxP0V+es6VPPSUhM6zk8iImp2jbV6GwjX4pap0JFbUQN65X1g==}
+    engines: {node: '>=18'}
+    cpu: [arm]
+    os: [android]
+
+  '@esbuild/android-x64@0.25.0':
+    resolution: {integrity: sha512-m/ix7SfKG5buCnxasr52+LI78SQ+wgdENi9CqyCXwjVR2X4Jkz+BpC3le3AoBPYTC9NHklwngVXvbJ9/Akhrfg==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [android]
+
+  '@esbuild/darwin-arm64@0.25.0':
+    resolution: {integrity: sha512-mVwdUb5SRkPayVadIOI78K7aAnPamoeFR2bT5nszFUZ9P8UpK4ratOdYbZZXYSqPKMHfS1wdHCJk1P1EZpRdvw==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [darwin]
+
+  '@esbuild/darwin-x64@0.25.0':
+    resolution: {integrity: sha512-DgDaYsPWFTS4S3nWpFcMn/33ZZwAAeAFKNHNa1QN0rI4pUjgqf0f7ONmXf6d22tqTY+H9FNdgeaAa+YIFUn2Rg==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [darwin]
+
+  '@esbuild/freebsd-arm64@0.25.0':
+    resolution: {integrity: sha512-VN4ocxy6dxefN1MepBx/iD1dH5K8qNtNe227I0mnTRjry8tj5MRk4zprLEdG8WPyAPb93/e4pSgi1SoHdgOa4w==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [freebsd]
+
+  '@esbuild/freebsd-x64@0.25.0':
+    resolution: {integrity: sha512-mrSgt7lCh07FY+hDD1TxiTyIHyttn6vnjesnPoVDNmDfOmggTLXRv8Id5fNZey1gl/V2dyVK1VXXqVsQIiAk+A==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [freebsd]
+
+  '@esbuild/linux-arm64@0.25.0':
+    resolution: {integrity: sha512-9QAQjTWNDM/Vk2bgBl17yWuZxZNQIF0OUUuPZRKoDtqF2k4EtYbpyiG5/Dk7nqeK6kIJWPYldkOcBqjXjrUlmg==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [linux]
+
+  '@esbuild/linux-arm@0.25.0':
+    resolution: {integrity: sha512-vkB3IYj2IDo3g9xX7HqhPYxVkNQe8qTK55fraQyTzTX/fxaDtXiEnavv9geOsonh2Fd2RMB+i5cbhu2zMNWJwg==}
+    engines: {node: '>=18'}
+    cpu: [arm]
+    os: [linux]
+
+  '@esbuild/linux-ia32@0.25.0':
+    resolution: {integrity: sha512-43ET5bHbphBegyeqLb7I1eYn2P/JYGNmzzdidq/w0T8E2SsYL1U6un2NFROFRg1JZLTzdCoRomg8Rvf9M6W6Gg==}
+    engines: {node: '>=18'}
+    cpu: [ia32]
+    os: [linux]
+
+  '@esbuild/linux-loong64@0.25.0':
+    resolution: {integrity: sha512-fC95c/xyNFueMhClxJmeRIj2yrSMdDfmqJnyOY4ZqsALkDrrKJfIg5NTMSzVBr5YW1jf+l7/cndBfP3MSDpoHw==}
+    engines: {node: '>=18'}
+    cpu: [loong64]
+    os: [linux]
+
+  '@esbuild/linux-mips64el@0.25.0':
+    resolution: {integrity: sha512-nkAMFju7KDW73T1DdH7glcyIptm95a7Le8irTQNO/qtkoyypZAnjchQgooFUDQhNAy4iu08N79W4T4pMBwhPwQ==}
+    engines: {node: '>=18'}
+    cpu: [mips64el]
+    os: [linux]
+
+  '@esbuild/linux-ppc64@0.25.0':
+    resolution: {integrity: sha512-NhyOejdhRGS8Iwv+KKR2zTq2PpysF9XqY+Zk77vQHqNbo/PwZCzB5/h7VGuREZm1fixhs4Q/qWRSi5zmAiO4Fw==}
+    engines: {node: '>=18'}
+    cpu: [ppc64]
+    os: [linux]
+
+  '@esbuild/linux-riscv64@0.25.0':
+    resolution: {integrity: sha512-5S/rbP5OY+GHLC5qXp1y/Mx//e92L1YDqkiBbO9TQOvuFXM+iDqUNG5XopAnXoRH3FjIUDkeGcY1cgNvnXp/kA==}
+    engines: {node: '>=18'}
+    cpu: [riscv64]
+    os: [linux]
+
+  '@esbuild/linux-s390x@0.25.0':
+    resolution: {integrity: sha512-XM2BFsEBz0Fw37V0zU4CXfcfuACMrppsMFKdYY2WuTS3yi8O1nFOhil/xhKTmE1nPmVyvQJjJivgDT+xh8pXJA==}
+    engines: {node: '>=18'}
+    cpu: [s390x]
+    os: [linux]
+
+  '@esbuild/linux-x64@0.25.0':
+    resolution: {integrity: sha512-9yl91rHw/cpwMCNytUDxwj2XjFpxML0y9HAOH9pNVQDpQrBxHy01Dx+vaMu0N1CKa/RzBD2hB4u//nfc+Sd3Cw==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [linux]
+
+  '@esbuild/netbsd-arm64@0.25.0':
+    resolution: {integrity: sha512-RuG4PSMPFfrkH6UwCAqBzauBWTygTvb1nxWasEJooGSJ/NwRw7b2HOwyRTQIU97Hq37l3npXoZGYMy3b3xYvPw==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [netbsd]
+
+  '@esbuild/netbsd-x64@0.25.0':
+    resolution: {integrity: sha512-jl+qisSB5jk01N5f7sPCsBENCOlPiS/xptD5yxOx2oqQfyourJwIKLRA2yqWdifj3owQZCL2sn6o08dBzZGQzA==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [netbsd]
+
+  '@esbuild/openbsd-arm64@0.25.0':
+    resolution: {integrity: sha512-21sUNbq2r84YE+SJDfaQRvdgznTD8Xc0oc3p3iW/a1EVWeNj/SdUCbm5U0itZPQYRuRTW20fPMWMpcrciH2EJw==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [openbsd]
+
+  '@esbuild/openbsd-x64@0.25.0':
+    resolution: {integrity: sha512-2gwwriSMPcCFRlPlKx3zLQhfN/2WjJ2NSlg5TKLQOJdV0mSxIcYNTMhk3H3ulL/cak+Xj0lY1Ym9ysDV1igceg==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [openbsd]
+
+  '@esbuild/sunos-x64@0.25.0':
+    resolution: {integrity: sha512-bxI7ThgLzPrPz484/S9jLlvUAHYMzy6I0XiU1ZMeAEOBcS0VePBFxh1JjTQt3Xiat5b6Oh4x7UC7IwKQKIJRIg==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [sunos]
+
+  '@esbuild/win32-arm64@0.25.0':
+    resolution: {integrity: sha512-ZUAc2YK6JW89xTbXvftxdnYy3m4iHIkDtK3CLce8wg8M2L+YZhIvO1DKpxrd0Yr59AeNNkTiic9YLf6FTtXWMw==}
+    engines: {node: '>=18'}
+    cpu: [arm64]
+    os: [win32]
+
+  '@esbuild/win32-ia32@0.25.0':
+    resolution: {integrity: sha512-eSNxISBu8XweVEWG31/JzjkIGbGIJN/TrRoiSVZwZ6pkC6VX4Im/WV2cz559/TXLcYbcrDN8JtKgd9DJVIo8GA==}
+    engines: {node: '>=18'}
+    cpu: [ia32]
+    os: [win32]
+
+  '@esbuild/win32-x64@0.25.0':
+    resolution: {integrity: sha512-ZENoHJBxA20C2zFzh6AI4fT6RraMzjYw4xKWemRTRmRVtN9c5DcH9r/f2ihEkMjOW5eGgrwCslG/+Y/3bL+DHQ==}
+    engines: {node: '>=18'}
+    cpu: [x64]
+    os: [win32]
+
+  '@protobufjs/aspromise@1.1.2':
+    resolution: {integrity: sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ==}
+
+  '@protobufjs/base64@1.1.2':
+    resolution: {integrity: sha512-AZkcAA5vnN/v4PDqKyMR5lx7hZttPDgClv83E//FMNhR2TMcLUhfRUBHCmSl0oi9zMgDDqRUJkSxO3wm85+XLg==}
+
+  '@protobufjs/codegen@2.0.4':
+    resolution: {integrity: sha512-YyFaikqM5sH0ziFZCN3xDC7zeGaB/d0IUb9CATugHWbd1FRFwWwt4ld4OYMPWu5a3Xe01mGAULCdqhMlPl29Jg==}
+
+  '@protobufjs/eventemitter@1.1.0':
+    resolution: {integrity: sha512-j9ednRT81vYJ9OfVuXG6ERSTdEL1xVsNgqpkxMsbIabzSo3goCjDIveeGv5d03om39ML71RdmrGNjG5SReBP/Q==}
+
+  '@protobufjs/fetch@1.1.0':
+    resolution: {integrity: sha512-lljVXpqXebpsijW71PZaCYeIcE5on1w5DlQy5WH6GLbFryLUrBD4932W/E2BSpfRJWseIL4v/KPgBFxDOIdKpQ==}
+
+  '@protobufjs/float@1.0.2':
+    resolution: {integrity: sha512-Ddb+kVXlXst9d+R9PfTIxh1EdNkgoRe5tOX6t01f1lYWOvJnSPDBlG241QLzcyPdoNTsblLUdujGSE4RzrTZGQ==}
+
+  '@protobufjs/inquire@1.1.0':
+    resolution: {integrity: sha512-kdSefcPdruJiFMVSbn801t4vFK7KB/5gd2fYvrxhuJYg8ILrmn9SKSX2tZdV6V+ksulWqS7aXjBcRXl3wHoD9Q==}
+
+  '@protobufjs/path@1.1.2':
+    resolution: {integrity: sha512-6JOcJ5Tm08dOHAbdR3GrvP+yUUfkjG5ePsHYczMFLq3ZmMkAD98cDgcT2iA1lJ9NVwFd4tH/iSSoe44YWkltEA==}
+
+  '@protobufjs/pool@1.1.0':
+    resolution: {integrity: sha512-0kELaGSIDBKvcgS4zkjz1PeddatrjYcmMWOlAuAPwAeccUrPHdUqo/J6LiymHHEiJT5NrF1UVwxY14f+fy4WQw==}
+
+  '@protobufjs/utf8@1.1.0':
+    resolution: {integrity: sha512-Vvn3zZrhQZkkBE8LSuW3em98c0FwgO4nxzv6OdSxPKJIEKY2bGbHn+mhGIPerzI4twdxaP8/0+06HBpwf345Lw==}
+
+  '@types/node@22.13.5':
+    resolution: {integrity: sha512-+lTU0PxZXn0Dr1NBtC7Y8cR21AJr87dLLU953CWA6pMxxv/UDc7jYAY90upcrie1nRcD6XNG5HOYEDtgW5TxAg==}
+
+  esbuild@0.25.0:
+    resolution: {integrity: sha512-BXq5mqc8ltbaN34cDqWuYKyNhX8D/Z0J1xdtdQ8UcIIIyJyz+ZMKUt58tF3SrZ85jcfN/PZYhjR5uDQAYNVbuw==}
+    engines: {node: '>=18'}
+    hasBin: true
+
+  flatbuffers@1.12.0:
+    resolution: {integrity: sha512-c7CZADjRcl6j0PlvFy0ZqXQ67qSEZfrVPynmnL+2zPc+NtMvrF8Y0QceMo7QqnSPc7+uWjUIAbvCQ5WIKlMVdQ==}
+
+  guid-typescript@1.0.9:
+    resolution: {integrity: sha512-Y8T4vYhEfwJOTbouREvG+3XDsjr8E3kIr7uf+JZ0BYloFsttiHU0WfvANVsR7TxNUJa/WpCnw/Ino/p+DeBhBQ==}
+
+  long@5.3.1:
+    resolution: {integrity: sha512-ka87Jz3gcx/I7Hal94xaN2tZEOPoUOEVftkQqZx2EeQRN7LGdfLlI3FvZ+7WDplm+vK2Urx9ULrvSowtdCieng==}
+
+  onnxruntime-common@1.20.1:
+    resolution: {integrity: sha512-YiU0s0IzYYC+gWvqD1HzLc46Du1sXpSiwzKb63PACIJr6LfL27VsXSXQvt68EzD3V0D5Bc0vyJTjmMxp0ylQiw==}
+
+  onnxruntime-web@1.20.1:
+    resolution: {integrity: sha512-TePF6XVpLL1rWVMIl5Y9ACBQcyCNFThZON/jgElNd9Txb73CIEGlklhYR3UEr1cp5r0rbGI6nDwwrs79g7WjoA==}
+
+  platform@1.3.6:
+    resolution: {integrity: sha512-fnWVljUchTro6RiCFvCXBbNhJc2NijN7oIQxbwsyL0buWJPG85v81ehlHI9fXrJsMNgTofEoWIQeClKpgxFLrg==}
+
+  protobufjs@7.4.0:
+    resolution: {integrity: sha512-mRUWCc3KUU4w1jU8sGxICXH/gNS94DvI1gxqDvBzhj1JpcsimQkYiOJfwsPUykUI5ZaspFbSgmBLER8IrQ3tqw==}
+    engines: {node: '>=12.0.0'}
+
+  typescript@5.8.3:
+    resolution: {integrity: sha512-p1diW6TqL9L07nNxvRMM7hMMw4c5XOo/1ibL4aAIGmSAt9slTE1Xgw5KWuof2uTOvCg9BY7ZRi+GaF+7sfgPeQ==}
+    engines: {node: '>=14.17'}
+    hasBin: true
+
+  undici-types@6.20.0:
+    resolution: {integrity: sha512-Ny6QZ2Nju20vw1SRHe3d9jVu6gJ+4e3+MMpqu7pqE5HT6WsTSlce++GQmK5UXS8mzV8DSYHrQH+Xrf2jVcuKNg==}
+
+snapshots:
+
+  '@biomejs/biome@1.9.4':
+    optionalDependencies:
+      '@biomejs/cli-darwin-arm64': 1.9.4
+      '@biomejs/cli-darwin-x64': 1.9.4
+      '@biomejs/cli-linux-arm64': 1.9.4
+      '@biomejs/cli-linux-arm64-musl': 1.9.4
+      '@biomejs/cli-linux-x64': 1.9.4
+      '@biomejs/cli-linux-x64-musl': 1.9.4
+      '@biomejs/cli-win32-arm64': 1.9.4
+      '@biomejs/cli-win32-x64': 1.9.4
+
+  '@biomejs/cli-darwin-arm64@1.9.4':
+    optional: true
+
+  '@biomejs/cli-darwin-x64@1.9.4':
+    optional: true
+
+  '@biomejs/cli-linux-arm64-musl@1.9.4':
+    optional: true
+
+  '@biomejs/cli-linux-arm64@1.9.4':
+    optional: true
+
+  '@biomejs/cli-linux-x64-musl@1.9.4':
+    optional: true
+
+  '@biomejs/cli-linux-x64@1.9.4':
+    optional: true
+
+  '@biomejs/cli-win32-arm64@1.9.4':
+    optional: true
+
+  '@biomejs/cli-win32-x64@1.9.4':
+    optional: true
+
+  '@esbuild/aix-ppc64@0.25.0':
+    optional: true
+
+  '@esbuild/android-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/android-arm@0.25.0':
+    optional: true
+
+  '@esbuild/android-x64@0.25.0':
+    optional: true
+
+  '@esbuild/darwin-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/darwin-x64@0.25.0':
+    optional: true
+
+  '@esbuild/freebsd-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/freebsd-x64@0.25.0':
+    optional: true
+
+  '@esbuild/linux-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/linux-arm@0.25.0':
+    optional: true
+
+  '@esbuild/linux-ia32@0.25.0':
+    optional: true
+
+  '@esbuild/linux-loong64@0.25.0':
+    optional: true
+
+  '@esbuild/linux-mips64el@0.25.0':
+    optional: true
+
+  '@esbuild/linux-ppc64@0.25.0':
+    optional: true
+
+  '@esbuild/linux-riscv64@0.25.0':
+    optional: true
+
+  '@esbuild/linux-s390x@0.25.0':
+    optional: true
+
+  '@esbuild/linux-x64@0.25.0':
+    optional: true
+
+  '@esbuild/netbsd-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/netbsd-x64@0.25.0':
+    optional: true
+
+  '@esbuild/openbsd-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/openbsd-x64@0.25.0':
+    optional: true
+
+  '@esbuild/sunos-x64@0.25.0':
+    optional: true
+
+  '@esbuild/win32-arm64@0.25.0':
+    optional: true
+
+  '@esbuild/win32-ia32@0.25.0':
+    optional: true
+
+  '@esbuild/win32-x64@0.25.0':
+    optional: true
+
+  '@protobufjs/aspromise@1.1.2': {}
+
+  '@protobufjs/base64@1.1.2': {}
+
+  '@protobufjs/codegen@2.0.4': {}
+
+  '@protobufjs/eventemitter@1.1.0': {}
+
+  '@protobufjs/fetch@1.1.0':
+    dependencies:
+      '@protobufjs/aspromise': 1.1.2
+      '@protobufjs/inquire': 1.1.0
+
+  '@protobufjs/float@1.0.2': {}
+
+  '@protobufjs/inquire@1.1.0': {}
+
+  '@protobufjs/path@1.1.2': {}
+
+  '@protobufjs/pool@1.1.0': {}
+
+  '@protobufjs/utf8@1.1.0': {}
+
+  '@types/node@22.13.5':
+    dependencies:
+      undici-types: 6.20.0
+
+  esbuild@0.25.0:
+    optionalDependencies:
+      '@esbuild/aix-ppc64': 0.25.0
+      '@esbuild/android-arm': 0.25.0
+      '@esbuild/android-arm64': 0.25.0
+      '@esbuild/android-x64': 0.25.0
+      '@esbuild/darwin-arm64': 0.25.0
+      '@esbuild/darwin-x64': 0.25.0
+      '@esbuild/freebsd-arm64': 0.25.0
+      '@esbuild/freebsd-x64': 0.25.0
+      '@esbuild/linux-arm': 0.25.0
+      '@esbuild/linux-arm64': 0.25.0
+      '@esbuild/linux-ia32': 0.25.0
+      '@esbuild/linux-loong64': 0.25.0
+      '@esbuild/linux-mips64el': 0.25.0
+      '@esbuild/linux-ppc64': 0.25.0
+      '@esbuild/linux-riscv64': 0.25.0
+      '@esbuild/linux-s390x': 0.25.0
+      '@esbuild/linux-x64': 0.25.0
+      '@esbuild/netbsd-arm64': 0.25.0
+      '@esbuild/netbsd-x64': 0.25.0
+      '@esbuild/openbsd-arm64': 0.25.0
+      '@esbuild/openbsd-x64': 0.25.0
+      '@esbuild/sunos-x64': 0.25.0
+      '@esbuild/win32-arm64': 0.25.0
+      '@esbuild/win32-ia32': 0.25.0
+      '@esbuild/win32-x64': 0.25.0
+
+  flatbuffers@1.12.0: {}
+
+  guid-typescript@1.0.9: {}
+
+  long@5.3.1: {}
+
+  onnxruntime-common@1.20.1: {}
+
+  onnxruntime-web@1.20.1:
+    dependencies:
+      flatbuffers: 1.12.0
+      guid-typescript: 1.0.9
+      long: 5.3.1
+      onnxruntime-common: 1.20.1
+      platform: 1.3.6
+      protobufjs: 7.4.0
+
+  platform@1.3.6: {}
+
+  protobufjs@7.4.0:
+    dependencies:
+      '@protobufjs/aspromise': 1.1.2
+      '@protobufjs/base64': 1.1.2
+      '@protobufjs/codegen': 2.0.4
+      '@protobufjs/eventemitter': 1.1.0
+      '@protobufjs/fetch': 1.1.0
+      '@protobufjs/float': 1.0.2
+      '@protobufjs/inquire': 1.1.0
+      '@protobufjs/path': 1.1.2
+      '@protobufjs/pool': 1.1.0
+      '@protobufjs/utf8': 1.1.0
+      '@types/node': 22.13.5
+      long: 5.3.1
+
+  typescript@5.8.3: {}
+
+  undici-types@6.20.0: {}

crates/sbv2_wasm/src-js/index.ts

@@ -0,0 +1,108 @@
+import * as wasm from "../pkg/sbv2_wasm.js";
+import { InferenceSession, Tensor } from "onnxruntime-web";
+
+export class ModelHolder {
+	private models: Map<string, [InferenceSession, wasm.StyleVectorWrap]> =
+		new Map();
+	constructor(
+		private tok: wasm.TokenizerWrap,
+		private deberta: InferenceSession,
+	) {}
+	public static async globalInit(buf: ArrayBufferLike) {
+		await wasm.default(buf);
+	}
+	public static async create(tok: string, deberta: ArrayBufferLike) {
+		return new ModelHolder(
+			wasm.load_tokenizer(tok),
+			await InferenceSession.create(deberta, {
+				executionProviders: ["webnn", "webgpu", "wasm", "cpu"],
+				graphOptimizationLevel: "all",
+			}),
+		);
+	}
+	public async synthesize(
+		name: string,
+		text: string,
+		style_id: number = 0,
+		style_weight: number = 1.0,
+		sdp_ratio: number = 0.4,
+		speed: number = 1.0,
+	) {
+		const mod = this.models.get(name);
+		if (!mod) throw new Error(`No model named ${name}`);
+		const [vits2, style] = mod;
+		return wasm.synthesize(
+			text,
+			this.tok,
+			async (a: BigInt64Array, b: BigInt64Array) => {
+				try {
+					const res = (
+						await this.deberta.run({
+							input_ids: new Tensor("int64", a, [1, a.length]),
+							attention_mask: new Tensor("int64", b, [1, b.length]),
+						})
+					)["output"];
+					return [new Uint32Array(res.dims), await res.getData(true)];
+				} catch (e) {
+					console.warn(e);
+					throw e;
+				}
+			},
+			async (
+				[a_shape, a_array]: any,
+				b_d: any,
+				c_d: any,
+				d_d: any,
+				e_d: any,
+				f: number,
+				g: number,
+			) => {
+				try {
+					const a = new Tensor("float32", a_array, [1, ...a_shape]);
+					const b = new Tensor("int64", b_d, [1, b_d.length]);
+					const c = new Tensor("int64", c_d, [1, c_d.length]);
+					const d = new Tensor("int64", d_d, [1, d_d.length]);
+					const e = new Tensor("float32", e_d, [1, e_d.length]);
+					const res = (
+						await vits2.run({
+							x_tst: b,
+							x_tst_lengths: new Tensor("int64", [b_d.length]),
+							sid: new Tensor("int64", [0]),
+							tones: c,
+							language: d,
+							bert: a,
+							style_vec: e,
+							sdp_ratio: new Tensor("float32", [f]),
+							length_scale: new Tensor("float32", [g]),
+							noise_scale: new Tensor("float32", [0.677]),
+							noise_scale_w: new Tensor("float32", [0.8]),
+						})
+					).output;
+					return [new Uint32Array(res.dims), await res.getData(true)];
+				} catch (e) {
+					console.warn(e);
+					throw e;
+				}
+			},
+			sdp_ratio,
+			1.0 / speed,
+			style_id,
+			style_weight,
+			style,
+		);
+	}
+	public async load(name: string, b: Uint8Array) {
+		const [style, vits2_b] = wasm.load_sbv2file(b);
+		const vits2 = await InferenceSession.create(vits2_b as Uint8Array, {
+			executionProviders: ["webnn", "webgpu", "wasm", "cpu"],
+			graphOptimizationLevel: "all",
+		});
+		this.models.set(name, [vits2, style]);
+	}
+	public async unload(name: string) {
+		return this.models.delete(name);
+	}
+	public modelList() {
+		return this.models.keys();
+	}
+}

crates/sbv2_wasm/src/array_helper.rs

@@ -0,0 +1,102 @@
+pub fn vec8_to_array8(v: Vec<u8>) -> js_sys::Uint8Array {
+    let arr = js_sys::Uint8Array::new_with_length(v.len() as u32);
+    arr.copy_from(&v);
+    arr
+}
+
+pub fn vec_f32_to_array_f32(v: Vec<f32>) -> js_sys::Float32Array {
+    let arr = js_sys::Float32Array::new_with_length(v.len() as u32);
+    arr.copy_from(&v);
+    arr
+}
+
+pub fn array8_to_vec8(buf: js_sys::Uint8Array) -> Vec<u8> {
+    let mut body = vec![0; buf.length() as usize];
+    buf.copy_to(&mut body[..]);
+    body
+}
+
+pub fn vec64_to_array64(v: Vec<i64>) -> js_sys::BigInt64Array {
+    let arr = js_sys::BigInt64Array::new_with_length(v.len() as u32);
+    arr.copy_from(&v);
+    arr
+}
+
+pub fn vec_to_array(v: Vec<wasm_bindgen::JsValue>) -> js_sys::Array {
+    let arr = js_sys::Array::new_with_length(v.len() as u32);
+    for (i, v) in v.into_iter().enumerate() {
+        arr.set(i as u32, v);
+    }
+    arr
+}
+
+struct A {
+    shape: Vec<u32>,
+    data: Vec<f32>,
+}
+
+impl TryFrom<wasm_bindgen::JsValue> for A {
+    type Error = sbv2_core::error::Error;
+
+    fn try_from(value: wasm_bindgen::JsValue) -> Result<Self, Self::Error> {
+        let value: js_sys::Array = value.into();
+        let mut shape = vec![];
+        let mut data = vec![];
+        for (i, v) in value.iter().enumerate() {
+            match i {
+                0 => {
+                    let v: js_sys::Uint32Array = v.into();
+                    shape = vec![0; v.length() as usize];
+                    v.copy_to(&mut shape);
+                }
+                1 => {
+                    let v: js_sys::Float32Array = v.into();
+                    data = vec![0.0; v.length() as usize];
+                    v.copy_to(&mut data);
+                }
+                _ => {}
+            };
+        }
+        Ok(A { shape, data })
+    }
+}
+
+pub fn array_to_array2_f32(
+    a: wasm_bindgen::JsValue,
+) -> sbv2_core::error::Result<ndarray::Array2<f32>> {
+    let a = A::try_from(a)?;
+    if a.shape.len() != 2 {
+        return Err(sbv2_core::error::Error::OtherError(
+            "Length mismatch".to_string(),
+        ));
+    }
+    let shape = [a.shape[0] as usize, a.shape[1] as usize];
+    let arr = ndarray::Array2::from_shape_vec(shape, a.data.to_vec())
+        .map_err(|e| sbv2_core::error::Error::OtherError(e.to_string()))?;
+    Ok(arr)
+}
+pub fn array_to_array3_f32(
+    a: wasm_bindgen::JsValue,
+) -> sbv2_core::error::Result<ndarray::Array3<f32>> {
+    let a = A::try_from(a)?;
+    if a.shape.len() != 3 {
+        return Err(sbv2_core::error::Error::OtherError(
+            "Length mismatch".to_string(),
+        ));
+    }
+    let shape = [
+        a.shape[0] as usize,
+        a.shape[1] as usize,
+        a.shape[2] as usize,
+    ];
+    let arr = ndarray::Array3::from_shape_vec(shape, a.data.to_vec())
+        .map_err(|e| sbv2_core::error::Error::OtherError(e.to_string()))?;
+    Ok(arr)
+}
+
+pub fn array2_f32_to_array(a: ndarray::Array2<f32>) -> js_sys::Array {
+    let shape: Vec<wasm_bindgen::JsValue> = a.shape().iter().map(|f| (*f as u32).into()).collect();
+    let typed_array = js_sys::Float32Array::new_with_length(a.len() as u32);
+    typed_array.copy_from(&a.into_flat().to_vec());
+    vec_to_array(vec![vec_to_array(shape).into(), typed_array.into()])
+}

crates/sbv2_wasm/src/lib.rs

@@ -0,0 +1,123 @@
+use once_cell::sync::Lazy;
+use sbv2_core::*;
+use wasm_bindgen::prelude::*;
+use wasm_bindgen_futures::JsFuture;
+mod array_helper;
+
+static JTALK: Lazy<jtalk::JTalk> = Lazy::new(|| jtalk::JTalk::new().unwrap());
+
+#[wasm_bindgen]
+pub struct TokenizerWrap {
+    tokenizer: tokenizer::Tokenizer,
+}
+
+#[wasm_bindgen]
+pub fn load_tokenizer(s: js_sys::JsString) -> Result<TokenizerWrap, JsError> {
+    if let Some(s) = s.as_string() {
+        Ok(TokenizerWrap {
+            tokenizer: tokenizer::Tokenizer::from_bytes(s.as_bytes())
+                .map_err(|e| JsError::new(&e.to_string()))?,
+        })
+    } else {
+        Err(JsError::new("invalid utf8"))
+    }
+}
+
+#[wasm_bindgen]
+pub struct StyleVectorWrap {
+    style_vector: ndarray::Array2<f32>,
+}
+
+#[wasm_bindgen]
+pub fn load_sbv2file(buf: js_sys::Uint8Array) -> Result<js_sys::Array, JsError> {
+    let (style_vectors, vits2) = sbv2file::parse_sbv2file(array_helper::array8_to_vec8(buf))?;
+    let buf = array_helper::vec8_to_array8(vits2);
+    Ok(array_helper::vec_to_array(vec![
+        StyleVectorWrap {
+            style_vector: style::load_style(style_vectors)?,
+        }
+        .into(),
+        buf.into(),
+    ]))
+}
+
+#[allow(clippy::too_many_arguments)]
+#[wasm_bindgen]
+pub async fn synthesize(
+    text: &str,
+    tokenizer: &TokenizerWrap,
+    bert_predict_fn: js_sys::Function,
+    synthesize_fn: js_sys::Function,
+    sdp_ratio: f32,
+    length_scale: f32,
+    style_id: i32,
+    style_weight: f32,
+    style_vectors: &StyleVectorWrap,
+) -> Result<js_sys::Uint8Array, JsError> {
+    let synthesize_wrap = |bert_ori: ndarray::Array2<f32>,
+                           x_tst: ndarray::Array1<i64>,
+                           tones: ndarray::Array1<i64>,
+                           lang_ids: ndarray::Array1<i64>,
+                           style_vector: ndarray::Array1<f32>,
+                           sdp_ratio: f32,
+                           length_scale: f32| async move {
+        let arr = array_helper::vec_to_array(vec![
+            array_helper::array2_f32_to_array(bert_ori).into(),
+            array_helper::vec64_to_array64(x_tst.to_vec()).into(),
+            array_helper::vec64_to_array64(tones.to_vec()).into(),
+            array_helper::vec64_to_array64(lang_ids.to_vec()).into(),
+            array_helper::vec_f32_to_array_f32(style_vector.to_vec()).into(),
+            sdp_ratio.into(),
+            length_scale.into(),
+        ]);
+        let res = synthesize_fn
+            .apply(&js_sys::Object::new().into(), &arr)
+            .map_err(|e| {
+                error::Error::OtherError(e.as_string().unwrap_or("unknown".to_string()))
+            })?;
+        let res = JsFuture::from(Into::<js_sys::Promise>::into(res))
+            .await
+            .map_err(|e| {
+                sbv2_core::error::Error::OtherError(e.as_string().unwrap_or("unknown".to_string()))
+            })?;
+        array_helper::array_to_array3_f32(res)
+    };
+    let (bert_ori, phones, tones, lang_ids) = tts_util::parse_text(
+        text,
+        &JTALK,
+        &tokenizer.tokenizer,
+        |token_ids: Vec<i64>, attention_masks: Vec<i64>| {
+            Box::pin(async move {
+                let arr = array_helper::vec_to_array(vec![
+                    array_helper::vec64_to_array64(token_ids).into(),
+                    array_helper::vec64_to_array64(attention_masks).into(),
+                ]);
+                let res = bert_predict_fn
+                    .apply(&js_sys::Object::new().into(), &arr)
+                    .map_err(|e| {
+                        error::Error::OtherError(e.as_string().unwrap_or("unknown".to_string()))
+                    })?;
+                let res = JsFuture::from(Into::<js_sys::Promise>::into(res))
+                    .await
+                    .map_err(|e| {
+                        sbv2_core::error::Error::OtherError(
+                            e.as_string().unwrap_or("unknown".to_string()),
+                        )
+                    })?;
+                array_helper::array_to_array2_f32(res)
+            })
+        },
+    )
+    .await?;
+    let audio = synthesize_wrap(
+        bert_ori.to_owned(),
+        phones,
+        tones,
+        lang_ids,
+        style::get_style_vector(&style_vectors.style_vector, style_id, style_weight)?,
+        sdp_ratio,
+        length_scale,
+    )
+    .await?;
+    Ok(array_helper::vec8_to_array8(tts_util::array_to_vec(audio)?))
+}

crates/sbv2_wasm/tsconfig.json

@@ -0,0 +1,15 @@
+{
+	"compilerOptions": {
+		"target": "ESNext",
+		"module": "ESNext",
+		"rootDir": "./src-js",
+		"outDir": "./dist",
+		"moduleResolution": "node",
+		"esModuleInterop": true,
+		"forceConsistentCasingInFileNames": true,
+		"strict": true,
+		"skipLibCheck": true,
+		"declaration": true,
+		"emitDeclarationOnly": true
+	}
+}

docker/cpu.Dockerfile

@@ -1,9 +0,0 @@
-FROM rust AS builder
-WORKDIR /work
-COPY . .
-RUN cargo build -r --bin sbv2_api
-FROM gcr.io/distroless/cc-debian12
-WORKDIR /work
-COPY --from=builder /work/target/release/sbv2_api /work/main
-COPY --from=builder /work/target/release/*.so /work
-CMD ["/work/main"]

sbv2_api/Cargo.toml

@@ -1,22 +0,0 @@
-[package]
-name = "sbv2_api"
-version = "0.1.0"
-edition = "2021"
-
-[dependencies]
-anyhow.workspace = true
-axum = "0.7.5"
-dotenvy.workspace = true
-env_logger.workspace = true
-log = "0.4.22"
-sbv2_core = { version = "0.1.3", path = "../sbv2_core" }
-serde = { version = "1.0.210", features = ["derive"] }
-tokio = { version = "1.40.0", features = ["full"] }
-
-[features]
-coreml = ["sbv2_core/coreml"]
-cuda = ["sbv2_core/cuda"]
-cuda_tf32 = ["sbv2_core/cuda_tf32"]
-dynamic = ["sbv2_core/dynamic"]
-directml = ["sbv2_core/directml"]
-tensorrt = ["sbv2_core/tensorrt"]

sbv2_bindings/Cargo.toml

@@ -1,15 +0,0 @@
-[package]
-name = "sbv2_bindings"
-version = "0.1.0"
-edition = "2021"
-
-# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
-[lib]
-name = "sbv2_bindings"
-crate-type = ["cdylib"]
-
-[dependencies]
-anyhow.workspace = true
-ndarray.workspace = true
-pyo3 = { version = "0.22.0", features = ["anyhow"] }
-sbv2_core = { version = "0.1.4", path = "../sbv2_core" }

sbv2_core/Cargo.toml

@@ -1,35 +0,0 @@
-[package]
-name = "sbv2_core"
-description = "Style-Bert-VITSの推論ライブラリ"
-version = "0.1.4"
-edition = "2021"
-license = "MIT"
-readme = "../README.md"
-repository = "https://github.com/tuna2134/sbv2-api"
-documentation = "https://docs.rs/sbv2_core"
-
-[dependencies]
-anyhow.workspace = true
-dotenvy.workspace = true
-env_logger.workspace = true
-hound = "3.5.1"
-jpreprocess = { version = "0.10.0", features = ["naist-jdic"] }
-ndarray.workspace = true
-num_cpus = "1.16.0"
-once_cell = "1.19.0"
-ort = { git = "https://github.com/pykeio/ort.git", version = "2.0.0-rc.6" }
-regex = "1.10.6"
-serde = { version = "1.0.210", features = ["derive"] }
-serde_json = "1.0.128"
-tar = "0.4.41"
-thiserror = "1.0.63"
-tokenizers = "0.20.0"
-zstd = "0.13.2"
-
-[features]
-cuda = ["ort/cuda"]
-cuda_tf32 = []
-dynamic = ["ort/load-dynamic"]
-directml = ["ort/directml"]
-tensorrt = ["ort/tensorrt"]
-coreml = ["ort/coreml"]

sbv2_core/src/bert.rs

@@ -1,23 +0,0 @@
-use crate::error::Result;
-use ndarray::Array2;
-use ort::Session;
-
-pub fn predict(
-    session: &Session,
-    token_ids: Vec<i64>,
-    attention_masks: Vec<i64>,
-) -> Result<Array2<f32>> {
-    let outputs = session.run(
-        ort::inputs! {
-            "input_ids" => Array2::from_shape_vec((1, token_ids.len()), token_ids).unwrap(),
-            "attention_mask" => Array2::from_shape_vec((1, attention_masks.len()), attention_masks).unwrap(),
-        }?
-    )?;
-
-    let output = outputs.get("output").unwrap();
-
-    let content = output.try_extract_tensor::<f32>()?.to_owned();
-    let (data, _) = content.clone().into_raw_vec_and_offset();
-
-    Ok(Array2::from_shape_vec((content.shape()[0], content.shape()[1]), data).unwrap())
-}

sbv2_core/src/jtalk.rs

@@ -1,403 +0,0 @@
-use crate::error::{Error, Result};
-use crate::mora::{MORA_KATA_TO_MORA_PHONEMES, VOWELS};
-use crate::norm::{replace_punctuation, PUNCTUATIONS};
-use jpreprocess::*;
-use once_cell::sync::Lazy;
-use regex::Regex;
-use std::cmp::Reverse;
-use std::collections::HashSet;
-use std::sync::Arc;
-
-type JPreprocessType = JPreprocess<DefaultFetcher>;
-
-fn initialize_jtalk() -> Result<JPreprocessType> {
-    let config = JPreprocessConfig {
-        dictionary: SystemDictionaryConfig::Bundled(kind::JPreprocessDictionaryKind::NaistJdic),
-        user_dictionary: None,
-    };
-    let jpreprocess = JPreprocess::from_config(config)?;
-    Ok(jpreprocess)
-}
-
-static JTALK_G2P_G_A1_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"/A:([0-9\-]+)\+").unwrap());
-static JTALK_G2P_G_A2_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"\+(\d+)\+").unwrap());
-static JTALK_G2P_G_A3_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"\+(\d+)/").unwrap());
-static JTALK_G2P_G_E3_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"!(\d+)_").unwrap());
-static JTALK_G2P_G_F1_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"/F:(\d+)_").unwrap());
-static JTALK_G2P_G_P3_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"\-(.*?)\+").unwrap());
-
-fn numeric_feature_by_regex(regex: &Regex, text: &str) -> i32 {
-    if let Some(mat) = regex.captures(text) {
-        mat[1].parse::<i32>().unwrap()
-    } else {
-        -50
-    }
-}
-
-macro_rules! hash_set {
-    ($($elem:expr),* $(,)?) => {{
-        let mut set = HashSet::new();
-        $(
-            set.insert($elem);
-        )*
-        set
-    }};
-}
-
-pub struct JTalk {
-    pub jpreprocess: Arc<JPreprocessType>,
-}
-
-impl JTalk {
-    pub fn new() -> Result<Self> {
-        let jpreprocess = Arc::new(initialize_jtalk()?);
-        Ok(Self { jpreprocess })
-    }
-
-    pub fn num2word(&self, text: &str) -> Result<String> {
-        let mut parsed = self.jpreprocess.text_to_njd(text)?;
-        parsed.preprocess();
-        let texts: Vec<String> = parsed
-            .nodes
-            .iter()
-            .map(|x| x.get_string().to_string())
-            .collect();
-        Ok(texts.join(""))
-    }
-
-    pub fn process_text(&self, text: &str) -> Result<JTalkProcess> {
-        let parsed = self.jpreprocess.run_frontend(text)?;
-        let jtalk_process = JTalkProcess::new(Arc::clone(&self.jpreprocess), parsed);
-        Ok(jtalk_process)
-    }
-}
-
-static KATAKANA_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"[\u30A0-\u30FF]+").unwrap());
-static MORA_PATTERN: Lazy<Vec<String>> = Lazy::new(|| {
-    let mut sorted_keys: Vec<String> = MORA_KATA_TO_MORA_PHONEMES.keys().cloned().collect();
-    sorted_keys.sort_by_key(|b| Reverse(b.len()));
-    sorted_keys
-});
-static LONG_PATTERN: Lazy<Regex> = Lazy::new(|| Regex::new(r"(\w)(ー*)").unwrap());
-
-pub struct JTalkProcess {
-    jpreprocess: Arc<JPreprocessType>,
-    parsed: Vec<String>,
-}
-
-impl JTalkProcess {
-    fn new(jpreprocess: Arc<JPreprocessType>, parsed: Vec<String>) -> Self {
-        Self {
-            jpreprocess,
-            parsed,
-        }
-    }
-
-    fn fix_phone_tone(&self, phone_tone_list: Vec<(String, i32)>) -> Result<Vec<(String, i32)>> {
-        let tone_values: HashSet<i32> = phone_tone_list
-            .iter()
-            .map(|(_letter, tone)| *tone)
-            .collect();
-        if tone_values.len() == 1 {
-            assert!(tone_values == hash_set![0], "{:?}", tone_values);
-            Ok(phone_tone_list)
-        } else if tone_values.len() == 2 {
-            if tone_values == hash_set![0, 1] {
-                return Ok(phone_tone_list);
-            } else if tone_values == hash_set![-1, 0] {
-                return Ok(phone_tone_list
-                    .iter()
-                    .map(|x| {
-                        let new_tone = if x.1 == -1 { 0 } else { 1 };
-                        (x.0.clone(), new_tone)
-                    })
-                    .collect());
-            } else {
-                return Err(Error::ValueError("Invalid tone values 0".to_string()));
-            }
-        } else {
-            return Err(Error::ValueError("Invalid tone values 1".to_string()));
-        }
-    }
-
-    pub fn g2p(&self) -> Result<(Vec<String>, Vec<i32>, Vec<i32>)> {
-        let phone_tone_list_wo_punct = self.g2phone_tone_wo_punct()?;
-        let (seq_text, seq_kata) = self.text_to_seq_kata()?;
-        let sep_phonemes = JTalkProcess::handle_long(
-            seq_kata
-                .iter()
-                .map(|x| JTalkProcess::kata_to_phoneme_list(x.clone()).unwrap())
-                .collect(),
-        );
-        let phone_w_punct: Vec<String> = sep_phonemes
-            .iter()
-            .flat_map(|x| x.iter())
-            .cloned()
-            .collect();
-
-        let mut phone_tone_list =
-            JTalkProcess::align_tones(phone_w_punct, phone_tone_list_wo_punct)?;
-
-        let mut sep_tokenized: Vec<Vec<String>> = Vec::new();
-        for seq_text_item in &seq_text {
-            let text = seq_text_item.clone();
-            if !PUNCTUATIONS.contains(&text.as_str()) {
-                sep_tokenized.push(text.chars().map(|x| x.to_string()).collect());
-            } else {
-                sep_tokenized.push(vec![text]);
-            }
-        }
-
-        let mut word2ph = Vec::new();
-        for (token, phoneme) in sep_tokenized.iter().zip(sep_phonemes.iter()) {
-            let phone_len = phoneme.len() as i32;
-            let word_len = token.len() as i32;
-            word2ph.append(&mut JTalkProcess::distribute_phone(phone_len, word_len));
-        }
-
-        let mut new_phone_tone_list = vec![("_".to_string(), 0)];
-        new_phone_tone_list.append(&mut phone_tone_list);
-        new_phone_tone_list.push(("_".to_string(), 0));
-
-        let mut new_word2ph = vec![1];
-        new_word2ph.extend(word2ph.clone());
-        new_word2ph.push(1);
-
-        let phones: Vec<String> = new_phone_tone_list.iter().map(|(x, _)| x.clone()).collect();
-        let tones: Vec<i32> = new_phone_tone_list.iter().map(|(_, x)| *x).collect();
-
-        Ok((phones, tones, new_word2ph))
-    }
-
-    fn distribute_phone(n_phone: i32, n_word: i32) -> Vec<i32> {
-        let mut phones_per_word = vec![0; n_word as usize];
-        for _ in 0..n_phone {
-            let min_task = phones_per_word.iter().min().unwrap();
-            let min_index = phones_per_word
-                .iter()
-                .position(|&x| x == *min_task)
-                .unwrap();
-            phones_per_word[min_index] += 1;
-        }
-        phones_per_word
-    }
-
-    fn align_tones(
-        phone_with_punct: Vec<String>,
-        phone_tone_list: Vec<(String, i32)>,
-    ) -> Result<Vec<(String, i32)>> {
-        let mut result: Vec<(String, i32)> = Vec::new();
-        let mut tone_index = 0;
-        for phone in phone_with_punct.clone() {
-            if tone_index >= phone_tone_list.len() {
-                result.push((phone, 0));
-            } else if phone == phone_tone_list[tone_index].0 {
-                result.push((phone, phone_tone_list[tone_index].1));
-                tone_index += 1;
-            } else if PUNCTUATIONS.contains(&phone.as_str()) {
-                result.push((phone, 0));
-            } else {
-                println!("phones {:?}", phone_with_punct);
-                println!("phone_tone_list: {:?}", phone_tone_list);
-                println!("result: {:?}", result);
-                println!("tone_index: {:?}", tone_index);
-                println!("phone: {:?}", phone);
-                return Err(Error::ValueError(format!("Mismatched phoneme: {}", phone)));
-            }
-        }
-
-        Ok(result)
-    }
-
-    fn handle_long(mut sep_phonemes: Vec<Vec<String>>) -> Vec<Vec<String>> {
-        for i in 0..sep_phonemes.len() {
-            if sep_phonemes[i].is_empty() {
-                continue;
-            }
-            if sep_phonemes[i][0] == "ー" {
-                if i != 0 {
-                    let prev_phoneme = sep_phonemes[i - 1].last().unwrap();
-                    if VOWELS.contains(&prev_phoneme.as_str()) {
-                        sep_phonemes[i][0] = prev_phoneme.clone();
-                    } else {
-                        sep_phonemes[i][0] = "ー".to_string();
-                    }
-                } else {
-                    sep_phonemes[i][0] = "ー".to_string();
-                }
-            }
-            if sep_phonemes[i].contains(&"ー".to_string()) {
-                for e in 0..sep_phonemes[i].len() {
-                    if sep_phonemes[i][e] == "ー" {
-                        sep_phonemes[i][e] =
-                            sep_phonemes[i][e - 1].chars().last().unwrap().to_string();
-                    }
-                }
-            }
-        }
-        sep_phonemes
-    }
-
-    fn kata_to_phoneme_list(mut text: String) -> Result<Vec<String>> {
-        if PUNCTUATIONS.contains(&text.as_str()) {
-            return Ok(text.chars().map(|x| x.to_string()).collect());
-        }
-        if !KATAKANA_PATTERN.is_match(&text) {
-            return Err(Error::ValueError(format!(
-                "Input must be katakana only: {}",
-                text
-            )));
-        }
-
-        for mora in MORA_PATTERN.iter() {
-            let mora = mora.to_string();
-            let (consonant, vowel) = MORA_KATA_TO_MORA_PHONEMES.get(&mora).unwrap();
-            if consonant.is_none() {
-                text = text.replace(&mora, &format!(" {}", vowel));
-            } else {
-                text = text.replace(
-                    &mora,
-                    &format!(" {} {}", consonant.as_ref().unwrap(), vowel),
-                );
-            }
-        }
-
-        let long_replacement = |m: &regex::Captures| {
-            let result = m.get(1).unwrap().as_str().to_string();
-            let mut second = String::new();
-            for _ in 0..m.get(2).unwrap().as_str().char_indices().count() {
-                second += &format!(" {}", m.get(1).unwrap().as_str());
-            }
-            result + &second
-        };
-        text = LONG_PATTERN
-            .replace_all(&text, long_replacement)
-            .to_string();
-
-        let data = text.trim().split(' ').map(|x| x.to_string()).collect();
-
-        Ok(data)
-    }
-
-    pub fn text_to_seq_kata(&self) -> Result<(Vec<String>, Vec<String>)> {
-        let mut seq_kata = vec![];
-        let mut seq_text = vec![];
-
-        for parts in &self.parsed {
-            let (string, pron) = self.parse_to_string_and_pron(parts.clone());
-            let mut yomi = pron.replace('’', "");
-            let word = replace_punctuation(string);
-            assert!(!yomi.is_empty(), "Empty yomi: {}", word);
-            if yomi == "、" {
-                if !word
-                    .chars()
-                    .all(|x| PUNCTUATIONS.contains(&x.to_string().as_str()))
-                {
-                    yomi = "'".repeat(word.len());
-                } else {
-                    yomi = word.clone();
-                }
-            } else if yomi == "？" {
-                assert!(word == "?", "yomi `？` comes from: {}", word);
-                yomi = "?".to_string();
-            }
-            seq_text.push(word);
-            seq_kata.push(yomi);
-        }
-        Ok((seq_text, seq_kata))
-    }
-
-    fn parse_to_string_and_pron(&self, parts: String) -> (String, String) {
-        let part_lists: Vec<String> = parts.split(',').map(|x| x.to_string()).collect();
-        (part_lists[0].clone(), part_lists[9].clone())
-    }
-
-    fn g2phone_tone_wo_punct(&self) -> Result<Vec<(String, i32)>> {
-        let prosodies = self.g2p_prosody()?;
-
-        let mut results: Vec<(String, i32)> = Vec::new();
-        let mut current_phrase: Vec<(String, i32)> = Vec::new();
-        let mut current_tone = 0;
-
-        for (i, letter) in prosodies.iter().enumerate() {
-            if letter == "^" {
-                assert!(i == 0);
-            } else if ["$", "?", "_", "#"].contains(&letter.as_str()) {
-                results.extend(self.fix_phone_tone(current_phrase.clone())?);
-                if ["$", "?"].contains(&letter.as_str()) {
-                    assert!(i == prosodies.len() - 1);
-                }
-                current_phrase = Vec::new();
-                current_tone = 0;
-            } else if letter == "[" {
-                current_tone += 1;
-            } else if letter == "]" {
-                current_tone -= 1;
-            } else {
-                let new_letter = if letter == "cl" {
-                    "q".to_string()
-                } else {
-                    letter.clone()
-                };
-                current_phrase.push((new_letter, current_tone));
-            }
-        }
-
-        Ok(results)
-    }
-
-    fn g2p_prosody(&self) -> Result<Vec<String>> {
-        let labels = self.jpreprocess.make_label(self.parsed.clone());
-
-        let mut phones: Vec<String> = Vec::new();
-        for (i, label) in labels.iter().enumerate() {
-            let mut p3 = {
-                let label_text = label.to_string();
-                let mattched = JTALK_G2P_G_P3_PATTERN.captures(&label_text).unwrap();
-                mattched[1].to_string()
-            };
-            if "AIUEO".contains(&p3) {
-                // 文字をlowerする
-                p3 = p3.to_lowercase();
-            }
-            if p3 == "sil" {
-                assert!(i == 0 || i == labels.len() - 1);
-                if i == 0 {
-                    phones.push("^".to_string());
-                } else if i == labels.len() - 1 {
-                    let e3 = numeric_feature_by_regex(&JTALK_G2P_G_E3_PATTERN, &label.to_string());
-                    if e3 == 0 {
-                        phones.push("$".to_string());
-                    } else if e3 == 1 {
-                        phones.push("?".to_string());
-                    }
-                }
-                continue;
-            } else if p3 == "pau" {
-                phones.push("_".to_string());
-                continue;
-            } else {
-                phones.push(p3.clone());
-            }
-
-            let a1 = numeric_feature_by_regex(&JTALK_G2P_G_A1_PATTERN, &label.to_string());
-            let a2 = numeric_feature_by_regex(&JTALK_G2P_G_A2_PATTERN, &label.to_string());
-            let a3 = numeric_feature_by_regex(&JTALK_G2P_G_A3_PATTERN, &label.to_string());
-
-            let f1 = numeric_feature_by_regex(&JTALK_G2P_G_F1_PATTERN, &label.to_string());
-
-            let a2_next =
-                numeric_feature_by_regex(&JTALK_G2P_G_A2_PATTERN, &labels[i + 1].to_string());
-
-            if a3 == 1 && a2_next == 1 && "aeiouAEIOUNcl".contains(&p3) {
-                phones.push("#".to_string());
-            } else if a1 == 0 && a2_next == a2 + 1 && a2 != f1 {
-                phones.push("]".to_string());
-            } else if a2 == 1 && a2_next == 2 {
-                phones.push("[".to_string());
-            }
-        }
-
-        Ok(phones)
-    }
-}

sbv2_core/src/main.rs

@@ -1,21 +0,0 @@
-use std::fs;
-
-use sbv2_core::tts;
-use std::env;
-
-fn main() -> anyhow::Result<()> {
-    dotenvy::dotenv_override().ok();
-    env_logger::init();
-    let text = fs::read_to_string("content.txt")?;
-    let ident = "aaa";
-    let mut tts_holder = tts::TTSModelHolder::new(
-        &fs::read(env::var("BERT_MODEL_PATH")?)?,
-        &fs::read(env::var("TOKENIZER_PATH")?)?,
-    )?;
-    tts_holder.load_sbv2file(ident, fs::read(env::var("MODEL_PATH")?)?)?;
-
-    let audio = tts_holder.easy_synthesize(ident, &text, 0, tts::SynthesizeOptions::default())?;
-    fs::write("output.wav", audio)?;
-
-    Ok(())
-}

sbv2_core/src/model.rs

@@ -1,93 +0,0 @@
-use crate::error::Result;
-use ndarray::{array, Array1, Array2, Array3, Axis};
-use ort::{GraphOptimizationLevel, Session};
-
-#[allow(clippy::vec_init_then_push, unused_variables)]
-pub fn load_model<P: AsRef<[u8]>>(model_file: P, bert: bool) -> Result<Session> {
-    let mut exp = Vec::new();
-    #[cfg(feature = "tensorrt")]
-    {
-        if bert {
-            exp.push(
-                ort::TensorRTExecutionProvider::default()
-                    .with_fp16(true)
-                    .with_profile_min_shapes("input_ids:1x1,attention_mask:1x1")
-                    .with_profile_max_shapes("input_ids:1x100,attention_mask:1x100")
-                    .with_profile_opt_shapes("input_ids:1x25,attention_mask:1x25")
-                    .build(),
-            );
-        }
-    }
-    #[cfg(feature = "cuda")]
-    {
-        #[allow(unused_mut)]
-        let mut cuda = ort::CUDAExecutionProvider::default()
-            .with_conv_algorithm_search(ort::CUDAExecutionProviderCuDNNConvAlgoSearch::Default);
-        #[cfg(feature = "cuda_tf32")]
-        {
-            cuda = cuda.with_tf32(true);
-        }
-        exp.push(cuda.build());
-    }
-    #[cfg(feature = "directml")]
-    {
-        exp.push(ort::DirectMLExecutionProvider::default().build());
-    }
-    #[cfg(feature = "coreml")]
-    {
-        exp.push(ort::CoreMLExecutionProvider::default().build());
-    }
-    exp.push(ort::CPUExecutionProvider::default().build());
-    Ok(Session::builder()?
-        .with_execution_providers(exp)?
-        .with_optimization_level(GraphOptimizationLevel::Level3)?
-        .with_intra_threads(num_cpus::get_physical())?
-        .with_parallel_execution(true)?
-        .with_inter_threads(num_cpus::get_physical())?
-        .commit_from_memory(model_file.as_ref())?)
-}
-
-#[allow(clippy::too_many_arguments)]
-pub fn synthesize(
-    session: &Session,
-    bert_ori: Array2<f32>,
-    x_tst: Array1<i64>,
-    tones: Array1<i64>,
-    lang_ids: Array1<i64>,
-    style_vector: Array1<f32>,
-    sdp_ratio: f32,
-    length_scale: f32,
-) -> Result<Array3<f32>> {
-    let bert = bert_ori.insert_axis(Axis(0));
-    let x_tst_lengths: Array1<i64> = array![x_tst.shape()[0] as i64];
-    let x_tst = x_tst.insert_axis(Axis(0));
-    let lang_ids = lang_ids.insert_axis(Axis(0));
-    let tones = tones.insert_axis(Axis(0));
-    let style_vector = style_vector.insert_axis(Axis(0));
-    let outputs = session.run(ort::inputs! {
-        "x_tst" => x_tst,
-        "x_tst_lengths" => x_tst_lengths,
-        "sid" => array![0_i64],
-        "tones" => tones,
-        "language" => lang_ids,
-        "bert" => bert,
-        "style_vec" => style_vector,
-        "sdp_ratio" => array![sdp_ratio],
-        "length_scale" => array![length_scale],
-    }?)?;
-
-    let audio_array = outputs
-        .get("output")
-        .unwrap()
-        .try_extract_tensor::<f32>()?
-        .to_owned();
-
-    Ok(Array3::from_shape_vec(
-        (
-            audio_array.shape()[0],
-            audio_array.shape()[1],
-            audio_array.shape()[2],
-        ),
-        audio_array.into_raw_vec_and_offset().0,
-    )?)
-}

sbv2_core/src/tts.rs

@@ -1,370 +0,0 @@
-use crate::error::{Error, Result};
-use crate::{bert, jtalk, model, nlp, norm, style, tokenizer, utils};
-use hound::{SampleFormat, WavSpec, WavWriter};
-use ndarray::{concatenate, s, Array, Array1, Array2, Array3, Axis};
-use ort::Session;
-use std::io::{Cursor, Read};
-use tar::Archive;
-use tokenizers::Tokenizer;
-use zstd::decode_all;
-
-#[derive(PartialEq, Eq, Clone)]
-pub struct TTSIdent(String);
-
-impl std::fmt::Display for TTSIdent {
-    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
-        f.write_str(&self.0)?;
-        Ok(())
-    }
-}
-
-impl<S> From<S> for TTSIdent
-where
-    S: AsRef<str>,
-{
-    fn from(value: S) -> Self {
-        TTSIdent(value.as_ref().to_string())
-    }
-}
-
-pub struct TTSModel {
-    vits2: Session,
-    style_vectors: Array2<f32>,
-    ident: TTSIdent,
-}
-
-/// High-level Style-Bert-VITS2's API
-pub struct TTSModelHolder {
-    tokenizer: Tokenizer,
-    bert: Session,
-    models: Vec<TTSModel>,
-    jtalk: jtalk::JTalk,
-}
-
-impl TTSModelHolder {
-    /// Initialize a new TTSModelHolder
-    ///
-    /// # Examples
-    ///
-    /// ```rs
-    /// let mut tts_holder = TTSModelHolder::new(std::fs::read("deberta.onnx")?, std::fs::read("tokenizer.json")?)?;
-    /// ```
-    pub fn new<P: AsRef<[u8]>>(bert_model_bytes: P, tokenizer_bytes: P) -> Result<Self> {
-        let bert = model::load_model(bert_model_bytes, true)?;
-        let jtalk = jtalk::JTalk::new()?;
-        let tokenizer = tokenizer::get_tokenizer(tokenizer_bytes)?;
-        Ok(TTSModelHolder {
-            bert,
-            models: vec![],
-            jtalk,
-            tokenizer,
-        })
-    }
-
-    /// Return a list of model names
-    pub fn models(&self) -> Vec<String> {
-        self.models.iter().map(|m| m.ident.to_string()).collect()
-    }
-
-    /// Load a .sbv2 file binary
-    ///
-    /// # Examples
-    ///
-    /// ```rs
-    /// tts_holder.load_sbv2file("tsukuyomi", std::fs::read("tsukuyomi.sbv2")?)?;
-    /// ```
-    pub fn load_sbv2file<I: Into<TTSIdent>, P: AsRef<[u8]>>(
-        &mut self,
-        ident: I,
-        sbv2_bytes: P,
-    ) -> Result<()> {
-        let mut arc = Archive::new(Cursor::new(decode_all(Cursor::new(sbv2_bytes.as_ref()))?));
-        let mut vits2 = None;
-        let mut style_vectors = None;
-        let mut et = arc.entries()?;
-        while let Some(Ok(mut e)) = et.next() {
-            let pth = String::from_utf8_lossy(&e.path_bytes()).to_string();
-            let mut b = Vec::with_capacity(e.size() as usize);
-            e.read_to_end(&mut b)?;
-            match pth.as_str() {
-                "model.onnx" => vits2 = Some(b),
-                "style_vectors.json" => style_vectors = Some(b),
-                _ => continue,
-            }
-        }
-        if style_vectors.is_none() {
-            return Err(Error::ModelNotFoundError("style_vectors".to_string()));
-        }
-        if vits2.is_none() {
-            return Err(Error::ModelNotFoundError("vits2".to_string()));
-        }
-        self.load(ident, style_vectors.unwrap(), vits2.unwrap())?;
-        Ok(())
-    }
-
-    /// Load a style vector and onnx model binary
-    ///
-    /// # Examples
-    ///
-    /// ```rs
-    /// tts_holder.load("tsukuyomi", std::fs::read("style_vectors.json")?, std::fs::read("model.onnx")?)?;
-    /// ```
-    pub fn load<I: Into<TTSIdent>, P: AsRef<[u8]>>(
-        &mut self,
-        ident: I,
-        style_vectors_bytes: P,
-        vits2_bytes: P,
-    ) -> Result<()> {
-        let ident = ident.into();
-        if self.find_model(ident.clone()).is_err() {
-            self.models.push(TTSModel {
-                vits2: model::load_model(vits2_bytes, false)?,
-                style_vectors: style::load_style(style_vectors_bytes)?,
-                ident,
-            })
-        }
-        Ok(())
-    }
-
-    /// Unload a model
-    pub fn unload<I: Into<TTSIdent>>(&mut self, ident: I) -> bool {
-        let ident = ident.into();
-        if let Some((i, _)) = self
-            .models
-            .iter()
-            .enumerate()
-            .find(|(_, m)| m.ident == ident)
-        {
-            self.models.remove(i);
-            true
-        } else {
-            false
-        }
-    }
-
-    /// Parse text and return the input for synthesize
-    ///
-    /// # Note
-    /// This function is for low-level usage, use `easy_synthesize` for high-level usage.
-    #[allow(clippy::type_complexity)]
-    pub fn parse_text(
-        &self,
-        text: &str,
-    ) -> Result<(Array2<f32>, Array1<i64>, Array1<i64>, Array1<i64>)> {
-        let text = self.jtalk.num2word(text)?;
-        let normalized_text = norm::normalize_text(&text);
-
-        let process = self.jtalk.process_text(&normalized_text)?;
-        let (phones, tones, mut word2ph) = process.g2p()?;
-        let (phones, tones, lang_ids) = nlp::cleaned_text_to_sequence(phones, tones);
-
-        let phones = utils::intersperse(&phones, 0);
-        let tones = utils::intersperse(&tones, 0);
-        let lang_ids = utils::intersperse(&lang_ids, 0);
-        for item in &mut word2ph {
-            *item *= 2;
-        }
-        word2ph[0] += 1;
-
-        let text = {
-            let (seq_text, _) = process.text_to_seq_kata()?;
-            seq_text.join("")
-        };
-        let (token_ids, attention_masks) = tokenizer::tokenize(&text, &self.tokenizer)?;
-
-        let bert_content = bert::predict(&self.bert, token_ids, attention_masks)?;
-
-        assert!(
-            word2ph.len() == text.chars().count() + 2,
-            "{} {}",
-            word2ph.len(),
-            normalized_text.chars().count()
-        );
-
-        let mut phone_level_feature = vec![];
-        for (i, reps) in word2ph.iter().enumerate() {
-            let repeat_feature = {
-                let (reps_rows, reps_cols) = (*reps, 1);
-                let arr_len = bert_content.slice(s![i, ..]).len();
-
-                let mut results: Array2<f32> =
-                    Array::zeros((reps_rows as usize, arr_len * reps_cols));
-
-                for j in 0..reps_rows {
-                    for k in 0..reps_cols {
-                        let mut view = results.slice_mut(s![j, k * arr_len..(k + 1) * arr_len]);
-                        view.assign(&bert_content.slice(s![i, ..]));
-                    }
-                }
-                results
-            };
-            phone_level_feature.push(repeat_feature);
-        }
-        let phone_level_feature = concatenate(
-            Axis(0),
-            &phone_level_feature
-                .iter()
-                .map(|x| x.view())
-                .collect::<Vec<_>>(),
-        )?;
-        let bert_ori = phone_level_feature.t();
-        Ok((
-            bert_ori.to_owned(),
-            phones.into(),
-            tones.into(),
-            lang_ids.into(),
-        ))
-    }
-
-    fn find_model<I: Into<TTSIdent>>(&self, ident: I) -> Result<&TTSModel> {
-        let ident = ident.into();
-        self.models
-            .iter()
-            .find(|m| m.ident == ident)
-            .ok_or(Error::ModelNotFoundError(ident.to_string()))
-    }
-
-    /// Get style vector by style id and weight
-    ///
-    /// # Note
-    /// This function is for low-level usage, use `easy_synthesize` for high-level usage.
-    pub fn get_style_vector<I: Into<TTSIdent>>(
-        &self,
-        ident: I,
-        style_id: i32,
-        weight: f32,
-    ) -> Result<Array1<f32>> {
-        style::get_style_vector(&self.find_model(ident)?.style_vectors, style_id, weight)
-    }
-
-    /// Synthesize text to audio
-    ///
-    /// # Examples
-    ///
-    /// ```rs
-    /// let audio = tts_holder.easy_synthesize("tsukuyomi", "こんにちは", 0, SynthesizeOptions::default())?;
-    /// ```
-    pub fn easy_synthesize<I: Into<TTSIdent> + Copy>(
-        &self,
-        ident: I,
-        text: &str,
-        style_id: i32,
-        options: SynthesizeOptions,
-    ) -> Result<Vec<u8>> {
-        let style_vector = self.get_style_vector(ident, style_id, options.style_weight)?;
-        let audio_array = if options.split_sentences {
-            let texts: Vec<&str> = text.split('\n').collect();
-            let mut audios = vec![];
-            for (i, t) in texts.iter().enumerate() {
-                if t.is_empty() {
-                    continue;
-                }
-                let (bert_ori, phones, tones, lang_ids) = self.parse_text(t)?;
-                let audio = model::synthesize(
-                    &self.find_model(ident)?.vits2,
-                    bert_ori.to_owned(),
-                    phones,
-                    tones,
-                    lang_ids,
-                    style_vector.clone(),
-                    options.sdp_ratio,
-                    options.length_scale,
-                )?;
-                audios.push(audio.clone());
-                if i != texts.len() - 1 {
-                    audios.push(Array3::zeros((1, 1, 22050)));
-                }
-            }
-            concatenate(
-                Axis(2),
-                &audios.iter().map(|x| x.view()).collect::<Vec<_>>(),
-            )?
-        } else {
-            let (bert_ori, phones, tones, lang_ids) = self.parse_text(text)?;
-            model::synthesize(
-                &self.find_model(ident)?.vits2,
-                bert_ori.to_owned(),
-                phones,
-                tones,
-                lang_ids,
-                style_vector,
-                options.sdp_ratio,
-                options.length_scale,
-            )?
-        };
-        Self::array_to_vec(audio_array)
-    }
-
-    fn array_to_vec(audio_array: Array3<f32>) -> Result<Vec<u8>> {
-        let spec = WavSpec {
-            channels: 1,
-            sample_rate: 44100,
-            bits_per_sample: 32,
-            sample_format: SampleFormat::Float,
-        };
-        let mut cursor = Cursor::new(Vec::new());
-        let mut writer = WavWriter::new(&mut cursor, spec)?;
-        for i in 0..audio_array.shape()[0] {
-            let output = audio_array.slice(s![i, 0, ..]).to_vec();
-            for sample in output {
-                writer.write_sample(sample)?;
-            }
-        }
-        writer.finalize()?;
-        Ok(cursor.into_inner())
-    }
-
-    /// Synthesize text to audio
-    ///
-    /// # Note
-    /// This function is for low-level usage, use `easy_synthesize` for high-level usage.
-    #[allow(clippy::too_many_arguments)]
-    pub fn synthesize<I: Into<TTSIdent>>(
-        &self,
-        ident: I,
-        bert_ori: Array2<f32>,
-        phones: Array1<i64>,
-        tones: Array1<i64>,
-        lang_ids: Array1<i64>,
-        style_vector: Array1<f32>,
-        sdp_ratio: f32,
-        length_scale: f32,
-    ) -> Result<Vec<u8>> {
-        let audio_array = model::synthesize(
-            &self.find_model(ident)?.vits2,
-            bert_ori.to_owned(),
-            phones,
-            tones,
-            lang_ids,
-            style_vector,
-            sdp_ratio,
-            length_scale,
-        )?;
-        Self::array_to_vec(audio_array)
-    }
-}
-
-/// Synthesize options
-///
-/// # Fields
-/// - `sdp_ratio`: SDP ratio
-/// - `length_scale`: Length scale
-/// - `style_weight`: Style weight
-/// - `split_sentences`: Split sentences
-pub struct SynthesizeOptions {
-    pub sdp_ratio: f32,
-    pub length_scale: f32,
-    pub style_weight: f32,
-    pub split_sentences: bool,
-}
-
-impl Default for SynthesizeOptions {
-    fn default() -> Self {
-        SynthesizeOptions {
-            sdp_ratio: 0.0,
-            length_scale: 1.0,
-            style_weight: 1.0,
-            split_sentences: true,
-        }
-    }
-}

scripts/.gitignore

@@ -0,0 +1,5 @@
+*.json
+venv/
+tmp/
+*.safetensors
+*.npy

scripts/convert/.python-version

@@ -0,0 +1 @@
+3.11

scripts/convert/convert_deberta.py

@@ -5,6 +5,7 @@ from transformers import AutoModelForMaskedLM, AutoTokenizer
 import torch
 from torch import nn
 from argparse import ArgumentParser
+import os
 
 parser = ArgumentParser()
 parser.add_argument("--model", default="ku-nlp/deberta-v2-large-japanese-char-wwm")
@@ -15,7 +16,7 @@ bert_models.load_tokenizer(Languages.JP, model_name)
 tokenizer = bert_models.load_tokenizer(Languages.JP)
 converter = BertConverter(tokenizer)
 tokenizer = converter.converted()
-tokenizer.save("../models/tokenizer.json")
+tokenizer.save("../../models/tokenizer.json")
 
 
 class ORTDeberta(nn.Module):
@@ -42,9 +43,10 @@ inputs = AutoTokenizer.from_pretrained(model_name)(
 torch.onnx.export(
     model,
     (inputs["input_ids"], inputs["token_type_ids"], inputs["attention_mask"]),
-    "../models/deberta.onnx",
+    "../../models/deberta.onnx",
     input_names=["input_ids", "token_type_ids", "attention_mask"],
     output_names=["output"],
     verbose=True,
     dynamic_axes={"input_ids": {1: "batch_size"}, "attention_mask": {1: "batch_size"}},
 )
+os.system("onnxsim ../../models/deberta.onnx ../../models/deberta.onnx")

scripts/convert/convert_model.py

@@ -36,7 +36,7 @@ data = array.tolist()
 hyper_parameters = HyperParameters.load_from_json(config_file)
 out_name = hyper_parameters.model_name
 
-with open(f"../models/style_vectors_{out_name}.json", "w") as f:
+with open(f"../../models/style_vectors_{out_name}.json", "w") as f:
     json.dump(
         {
             "data": data,
@@ -94,7 +94,7 @@ model = get_net_g(
 )
 
 
-def forward(x, x_len, sid, tone, lang, bert, style, length_scale, sdp_ratio):
+def forward(x, x_len, sid, tone, lang, bert, style, length_scale, sdp_ratio, noise_scale, noise_scale_w):
     return model.infer(
         x,
         x_len,
@@ -105,6 +105,8 @@ def forward(x, x_len, sid, tone, lang, bert, style, length_scale, sdp_ratio):
         style,
         sdp_ratio=sdp_ratio,
         length_scale=length_scale,
+        noise_scale=noise_scale,
+        noise_scale_w=noise_scale_w,
     )
 
 
@@ -122,8 +124,10 @@ torch.onnx.export(
         style_vec_tensor,
         torch.tensor(1.0),
         torch.tensor(0.0),
+        torch.tensor(0.6777),
+        torch.tensor(0.8),
     ),
-    f"../models/model_{out_name}.onnx",
+    f"../../models/model_{out_name}.onnx",
     verbose=True,
     dynamic_axes={
         "x_tst": {0: "batch_size", 1: "x_tst_max_length"},
@@ -144,14 +148,16 @@ torch.onnx.export(
         "style_vec",
         "length_scale",
         "sdp_ratio",
+        "noise_scale",
+        "noise_scale_w"
     ],
     output_names=["output"],
 )
-os.system(f"onnxsim ../models/model_{out_name}.onnx ../models/model_{out_name}.onnx")
-onnxfile = open(f"../models/model_{out_name}.onnx", "rb").read()
-stylefile = open(f"../models/style_vectors_{out_name}.json", "rb").read()
+os.system(f"onnxsim ../../models/model_{out_name}.onnx ../../models/model_{out_name}.onnx")
+onnxfile = open(f"../../models/model_{out_name}.onnx", "rb").read()
+stylefile = open(f"../../models/style_vectors_{out_name}.json", "rb").read()
 version = bytes("1", "utf8")
-with taropen(f"../models/tmp_{out_name}.sbv2tar", "w") as w:
+with taropen(f"../../models/tmp_{out_name}.sbv2tar", "w") as w:
 
     def add_tar(f, b):
         t = TarInfo(f)
@@ -161,9 +167,9 @@ with taropen(f"../models/tmp_{out_name}.sbv2tar", "w") as w:
     add_tar("version.txt", version)
     add_tar("model.onnx", onnxfile)
     add_tar("style_vectors.json", stylefile)
-open(f"../models/{out_name}.sbv2", "wb").write(
+open(f"../../models/{out_name}.sbv2", "wb").write(
     ZstdCompressor(threads=-1, level=22).compress(
-        open(f"../models/tmp_{out_name}.sbv2tar", "rb").read()
+        open(f"../../models/tmp_{out_name}.sbv2tar", "rb").read()
     )
 )
-os.unlink(f"../models/tmp_{out_name}.sbv2tar")
+os.unlink(f"../../models/tmp_{out_name}.sbv2tar")

scripts/convert/requirements.txt

@@ -0,0 +1,6 @@
+git+https://github.com/neodyland/style-bert-vits2-ref
+onnxsim
+numpy<2
+zstandard
+onnxruntime
+cmake<4

scripts/docker/cpu.Dockerfile

@@ -0,0 +1,17 @@
+FROM rust AS builder
+WORKDIR /work
+COPY . .
+RUN cargo build -r --bin sbv2_api
+FROM ubuntu AS upx
+WORKDIR /work
+RUN apt update && apt-get install -y upx binutils
+COPY --from=builder /work/target/release/sbv2_api /work/main
+COPY --from=builder /work/target/release/*.so /work
+RUN upx --best --lzma /work/main
+RUN find /work -maxdepth 1 -name "*.so" -exec strip --strip-unneeded {} +
+RUN find /work -maxdepth 1 -name "*.so" -exec upx --best --lzma {} +
+FROM gcr.io/distroless/cc-debian12
+WORKDIR /work
+COPY --from=upx /work/main /work/main
+COPY --from=upx /work/*.so /work
+CMD ["/work/main"]

scripts/docker/cuda.Dockerfile

@@ -2,9 +2,16 @@ FROM rust AS builder
 WORKDIR /work
 COPY . .
 RUN cargo build -r --bin sbv2_api -F cuda,cuda_tf32
-FROM nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04
+FROM ubuntu AS upx
 WORKDIR /work
+RUN apt update && apt-get install -y upx binutils
 COPY --from=builder /work/target/release/sbv2_api /work/main
 COPY --from=builder /work/target/release/*.so /work
+RUN upx --best --lzma /work/main
+RUN find /work -maxdepth 1 -name "*.so" -exec strip --strip-unneeded {} +
+FROM nvidia/cuda:12.3.2-cudnn9-runtime-ubuntu22.04
+WORKDIR /work
+COPY --from=upx /work/main /work/main
+COPY --from=upx /work/*.so /work
 ENV LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/work
-CMD ["/work/main"]
+CMD ["/work/main"]

scripts/docker/run_cpu.sh

@@ -1,3 +1,3 @@
 docker run -it --rm -p 3000:3000 --name sbv2 \
 -v ./models:/work/models --env-file .env \
-ghcr.io/tuna2134/sbv2-api:cpu
+ghcr.io/neodyland/sbv2-api:cpu

scripts/docker/run_cuda.sh

@@ -1,4 +1,4 @@
 docker run -it --rm -p 3000:3000 --name sbv2 \
 -v ./models:/work/models --env-file .env \
 --gpus all \
-ghcr.io/tuna2134/sbv2-api:cuda
+ghcr.io/neodyland/sbv2-api:cuda

scripts/make_dict.sh

@@ -0,0 +1,14 @@
+#!/bin/bash
+set -e
+git clone https://github.com/Aivis-Project/AivisSpeech-Engine ./scripts/tmp --filter=blob:none -n
+cd ./scripts/tmp
+git checkout 168b2a1144afe300b0490d9a6dd773ec6e927667 -- resources/dictionaries/*.csv
+cd ../..
+rm -rf ./crates/sbv2_core/src/dic
+cp -r ./scripts/tmp/resources/dictionaries ./crates/sbv2_core/src/dic
+rm -rf ./scripts/tmp
+for file in ./crates/sbv2_core/src/dic/0*.csv; do
+    /usr/bin/cat "$file"
+    echo
+done > ./crates/sbv2_core/src/all.csv
+lindera build ./crates/sbv2_core/src/all.csv ./crates/sbv2_core/src/dic/all.dic -u -k ipadic

scripts/sbv2-bindings-colab.ipynb

@@ -0,0 +1,180 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "# 音声合成プログラム\n",
+    "\n",
+    "このノートブックでは、`sbv2_bindings` パッケージを使用して音声合成を行います。必要なモデルをダウンロードし、ユーザーが入力したテキストから音声を生成します。音声合成が終わったら、再度テキストの入力を求め、ユーザーが終了するまで繰り返します。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 必要なパッケージのインストール\n",
+    "%pip install sbv2_bindings\n",
+    "\n",
+    "# 必要なモジュールのインポート\n",
+    "import os\n",
+    "import urllib.request\n",
+    "import time\n",
+    "from sbv2_bindings import TTSModel"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## モデルのダウンロード\n",
+    "\n",
+    "モデルファイルとトークナイザーをダウンロードします。ユーザーが独自のモデルを使用したい場合は、該当するURLまたはローカルパスを指定してください。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# モデルの URL またはローカルパスの指定\n",
+    "user_sbv2_model_url = \"\"  # カスタムモデルのURLがあればここに指定\n",
+    "user_sbv2_model_path = \"\"  # カスタムモデルのローカルパスがあればここに指定\n",
+    "\n",
+    "# モデル用のディレクトリを作成\n",
+    "model_dir = 'models'\n",
+    "os.makedirs(model_dir, exist_ok=True)\n",
+    "\n",
+    "# ダウンロードするファイルの URL\n",
+    "file_urls = [\n",
+    "    \"https://huggingface.co/googlefan/sbv2_onnx_models/resolve/main/tokenizer.json\",\n",
+    "    \"https://huggingface.co/googlefan/sbv2_onnx_models/resolve/main/deberta.onnx\",\n",
+    "]\n",
+    "\n",
+    "# モデルのパス決定\n",
+    "if user_sbv2_model_path:\n",
+    "    sbv2_model_path = user_sbv2_model_path  # ローカルモデルのパスを使用\n",
+    "elif user_sbv2_model_url:\n",
+    "    sbv2_model_filename = os.path.basename(user_sbv2_model_url)\n",
+    "    sbv2_model_path = os.path.join(model_dir, sbv2_model_filename)\n",
+    "    file_urls.append(user_sbv2_model_url)\n",
+    "else:\n",
+    "    # デフォルトのモデルを使用\n",
+    "    sbv2_model_filename = \"tsukuyomi.sbv2\"\n",
+    "    sbv2_model_path = os.path.join(model_dir, sbv2_model_filename)\n",
+    "    file_urls.append(\"https://huggingface.co/googlefan/sbv2_onnx_models/resolve/main/tsukuyomi.sbv2\")\n",
+    "\n",
+    "# ファイルをダウンロード\n",
+    "for url in file_urls:\n",
+    "    file_name = os.path.join(model_dir, os.path.basename(url))\n",
+    "    if not os.path.exists(file_name):\n",
+    "        print(f\"{file_name} をダウンロードしています...\")\n",
+    "        urllib.request.urlretrieve(url, file_name)\n",
+    "    else:\n",
+    "        print(f\"{file_name} は既に存在します。\")\n",
+    "\n",
+    "# ダウンロードまたは使用するファイルを確認\n",
+    "print(\"\\n使用するファイル:\")\n",
+    "for file in os.listdir(model_dir):\n",
+    "    print(file)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## モデルの読み込みと音声合成\n",
+    "\n",
+    "モデルを読み込み、ユーザーが入力したテキストから音声を生成します。話者名は使用する `.sbv2` ファイル名から自動的に取得します。音声合成が終わったら、再度テキストの入力を求め、ユーザーが終了するまで繰り返します。"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "# 音声合成の実行\n",
+    "def main():\n",
+    "    try:\n",
+    "        print(\"\\nモデルを読み込んでいます...\")\n",
+    "        model = TTSModel.from_path(\n",
+    "            os.path.join(model_dir, \"deberta.onnx\"),\n",
+    "            os.path.join(model_dir, \"tokenizer.json\")\n",
+    "        )\n",
+    "        print(\"モデルの読み込みが完了しました！\")\n",
+    "    except Exception as e:\n",
+    "        print(f\"モデルの読み込みに失敗しました: {e}\")\n",
+    "        return\n",
+    "\n",
+    "    # 話者名を取得（.sbv2 ファイル名の拡張子を除いた部分）\n",
+    "    speaker_name = os.path.splitext(os.path.basename(sbv2_model_path))[0]\n",
+    "    \n",
+    "    # 指定されたモデルのパスを使用\n",
+    "    try:\n",
+    "        model.load_sbv2file_from_path(speaker_name, sbv2_model_path)\n",
+    "        print(f\"話者 '{speaker_name}' のセットアップが完了しました！\")\n",
+    "    except Exception as e:\n",
+    "        print(f\"SBV2ファイルの読み込みに失敗しました: {e}\")\n",
+    "        return\n",
+    "\n",
+    "    # 音声合成を繰り返し実行\n",
+    "    while True:\n",
+    "        # 合成したいテキストをユーザーから入力\n",
+    "        user_input = input(\"\\n音声合成したいテキストを入力してください（終了するには 'exit' と入力）: \")\n",
+    "        \n",
+    "        if user_input.strip().lower() == 'exit':\n",
+    "            print(\"音声合成を終了します。\")\n",
+    "            break\n",
+    "\n",
+    "        # 出力ファイル名\n",
+    "        output_file = \"output.wav\"\n",
+    "\n",
+    "        # 音声合成を実行\n",
+    "        try:\n",
+    "            print(\"\\n音声合成を開始します...\")\n",
+    "            start_time = time.time()\n",
+    "\n",
+    "            audio_data = model.synthesize(user_input, speaker_name, 0, 0.0, 1)\n",
+    "\n",
+    "            with open(output_file, \"wb\") as f:\n",
+    "                f.write(audio_data)\n",
+    "\n",
+    "            end_time = time.time()\n",
+    "            elapsed_time = end_time - start_time\n",
+    "\n",
+    "            print(f\"\\n音声が '{output_file}' に保存されました。\")\n",
+    "            print(f\"音声合成にかかった時間: {elapsed_time:.2f} 秒\")\n",
+    "        except Exception as e:\n",
+    "            print(f\"音声合成に失敗しました: {e}\")\n",
+    "\n",
+    "if __name__ == \"__main__\":\n",
+    "    main()"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.x"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

scripts/sbv2-test-api.py

@@ -0,0 +1,8 @@
+import requests
+
+res = requests.post(
+    "http://localhost:3000/synthesize",
+    json={"text": "おはようございます", "ident": "tsukuyomi"},
+)
+with open("output.wav", "wb") as f:
+    f.write(res.content)

scripts/sbv2-test-bindings.py

@@ -3,16 +3,16 @@ from sbv2_bindings import TTSModel
 
 def main():
     print("Loading models...")
-    model = TTSModel.from_path("../models/debert.onnx", "../models/tokenizer.json")
+    model = TTSModel.from_path("./models/debert.onnx", "./models/tokenizer.json")
     print("Models loaded!")
 
-    model.load_sbv2file_from_path("amitaro", "../models/amitaro.sbv2")
+    model.load_sbv2file_from_path("amitaro", "./models/amitaro.sbv2")
     print("All setup is done!")
 
     style_vector = model.get_style_vector("amitaro", 0, 1.0)
     with open("output.wav", "wb") as f:
         f.write(
-            model.synthesize("おはようございます。", "amitaro", style_vector, 0.0, 0.5)
+            model.synthesize("おはようございます。", "amitaro", 0, 0, 0.0, 0.5)
         )
 
 

test.py

@@ -1,8 +1,19 @@
 import requests
 
-res = requests.post(
-    "http://localhost:3000/synthesize",
-    json={"text": "おはようございます", "ident": "tsukuyomi"},
-)
-with open("output.wav", "wb") as f:
-    f.write(res.content)
+
+data = (requests.get("http://localhost:8080/audio_query", params={
+    "text": "こんにちは、今日はいい天気ですね。",
+})).json()
+print(data)
+
+data = (requests.post("http://localhost:8080/synthesis", json={
+    "text": data["text"],
+    "ident": "tsukuyomi",
+    "speaker_id": 0,
+    "style_id": 0,
+    "sdp_ratio": 0.5,
+    "length_scale": 0.5,
+    "audio_query": data["audio_query"],
+})).content
+with open("test.wav", "wb") as f:
+    f.write(data)