From fc944b9d33c0fa58f95692cf2e89e44572b751ae Mon Sep 17 00:00:00 2001
From: Masato Kikuchi <m-kikuchi@coe.ad.jp>
Date: Wed, 26 Mar 2025 15:14:22 +0900
Subject: [PATCH] split the code for support voicevox

---
 Cargo.lock                       |  9 +++++++++
 Cargo.toml                       |  2 +-
 crates/sbv2_core/src/tts_util.rs | 19 ++++++++++++++-----
 crates/sbv2_voicevox/Cargo.toml  | 14 ++++++++++++++
 crates/sbv2_voicevox/README.md   |  2 ++
 crates/sbv2_voicevox/src/main.rs |  5 +++++
 6 files changed, 45 insertions(+), 6 deletions(-)
 create mode 100644 crates/sbv2_voicevox/Cargo.toml
 create mode 100644 crates/sbv2_voicevox/README.md
 create mode 100644 crates/sbv2_voicevox/src/main.rs

diff --git a/Cargo.lock b/Cargo.lock
index 3245a45..1d69f54 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -2312,6 +2312,15 @@ dependencies = [
  "zstd",
 ]
 
+[[package]]
+name = "sbv2_voicevox"
+version = "0.2.0-alpha6"
+dependencies = [
+ "anyhow",
+ "axum",
+ "sbv2_core",
+]
+
 [[package]]
 name = "sbv2_wasm"
 version = "0.2.0-alpha6"
diff --git a/Cargo.toml b/Cargo.toml
index 240b6e0..fcc17c0 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,6 +1,6 @@
 [workspace]
 resolver = "2"
-members = ["./crates/sbv2_api", "./crates/sbv2_core", "./crates/sbv2_bindings", "./crates/sbv2_wasm"]
+members = ["./crates/sbv2_api", "./crates/sbv2_core", "./crates/sbv2_bindings", "./crates/sbv2_wasm", "crates/sbv2_voicevox"]
 
 [workspace.package]
 version = "0.2.0-alpha6"
diff --git a/crates/sbv2_core/src/tts_util.rs b/crates/sbv2_core/src/tts_util.rs
index 24b059a..8cab20d 100644
--- a/crates/sbv2_core/src/tts_util.rs
+++ b/crates/sbv2_core/src/tts_util.rs
@@ -5,6 +5,19 @@ 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<(Vec<String>, Vec<i32>, Vec<i32>)> {
+    let text = jtalk.num2word(text)?;
+    let normalized_text = norm::normalize_text(&text);
+
+    let process = jtalk.process_text(&normalized_text)?;
+    let result = process.g2p()?;
+    Ok(result)
+}
+
 /// Parse text and return the input for synthesize
 ///
 /// # Note
@@ -21,11 +34,7 @@ pub async fn parse_text(
         Box<dyn std::future::Future<Output = 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, tones, mut word2ph) = process.g2p()?;
+    let (phones, tones, mut word2ph) = preprocess_parse_text(text, jtalk)?;
     let (phones, tones, lang_ids) = nlp::cleaned_text_to_sequence(phones, tones);
 
     let phones = utils::intersperse(&phones, 0);
diff --git a/crates/sbv2_voicevox/Cargo.toml b/crates/sbv2_voicevox/Cargo.toml
new file mode 100644
index 0000000..4bd10aa
--- /dev/null
+++ b/crates/sbv2_voicevox/Cargo.toml
@@ -0,0 +1,14 @@
+[package]
+name = "sbv2_voicevox"
+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"
+sbv2_core = { version = "0.2.0-alpha6", path = "../sbv2_core" }
diff --git a/crates/sbv2_voicevox/README.md b/crates/sbv2_voicevox/README.md
new file mode 100644
index 0000000..d9b4b1f
--- /dev/null
+++ b/crates/sbv2_voicevox/README.md
@@ -0,0 +1,2 @@
+# sbv2-voicevox
+sbv2-apiをvoicevox化します。
\ No newline at end of file
diff --git a/crates/sbv2_voicevox/src/main.rs b/crates/sbv2_voicevox/src/main.rs
new file mode 100644
index 0000000..8d6f05f
--- /dev/null
+++ b/crates/sbv2_voicevox/src/main.rs
@@ -0,0 +1,5 @@
+
+async fn main() -> anyhow::Result<()> {
+    println!("Hello, world!");
+    Ok(())
+}