From f4de3e15ae238a68be603235e2ce9f713ea41397 Mon Sep 17 00:00:00 2001
From: Masato Kikuchi <m-kikuchi@coe.ad.jp>
Date: Wed, 26 Mar 2025 16:14:29 +0900
Subject: [PATCH] initial commit: voicevox

---
 Cargo.lock                       |  5 +++--
 crates/sbv2_core/src/tts_util.rs | 10 ++++++----
 crates/sbv2_voicevox/Cargo.toml  |  1 +
 crates/sbv2_voicevox/src/main.rs |  7 ++++++-
 4 files changed, 16 insertions(+), 7 deletions(-)

diff --git a/Cargo.lock b/Cargo.lock
index 1d69f54..97f0509 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -2319,6 +2319,7 @@ dependencies = [
  "anyhow",
  "axum",
  "sbv2_core",
+ "tokio",
 ]
 
 [[package]]
@@ -2741,9 +2742,9 @@ dependencies = [
 
 [[package]]
 name = "tokio"
-version = "1.43.0"
+version = "1.44.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3d61fa4ffa3de412bfea335c6ecff681de2b609ba3c77ef3e00e521813a9ed9e"
+checksum = "f382da615b842244d4b8738c82ed1275e6c5dd90c459a30941cd07080b06c91a"
 dependencies = [
  "backtrace",
  "bytes",
diff --git a/crates/sbv2_core/src/tts_util.rs b/crates/sbv2_core/src/tts_util.rs
index 8cab20d..1334128 100644
--- a/crates/sbv2_core/src/tts_util.rs
+++ b/crates/sbv2_core/src/tts_util.rs
@@ -1,6 +1,7 @@
 use std::io::Cursor;
 
 use crate::error::Result;
+use crate::jtalk::JTalkProcess;
 use crate::{jtalk, nlp, norm, tokenizer, utils};
 use hound::{SampleFormat, WavSpec, WavWriter};
 use ndarray::{concatenate, s, Array, Array1, Array2, Array3, Axis};
@@ -9,13 +10,13 @@ use tokenizers::Tokenizer;
 pub fn preprocess_parse_text(
     text: &str,
     jtalk: &jtalk::JTalk,
-) -> Result<(Vec<String>, Vec<i32>, Vec<i32>)> {
+) -> Result<(Vec<String>, Vec<i32>, Vec<i32>, String, JTalkProcess)> {
     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)
+    let (phones, tones, word2ph) = process.g2p()?;
+    Ok((phones, tones, word2ph, normalized_text, process))
 }
 
 /// Parse text and return the input for synthesize
@@ -34,7 +35,8 @@ 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 (phones, tones, mut word2ph) = preprocess_parse_text(text, jtalk)?;
+    let (phones, tones, mut word2ph, normalized_text, process) =
+        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
index 4bd10aa..14c7b92 100644
--- a/crates/sbv2_voicevox/Cargo.toml
+++ b/crates/sbv2_voicevox/Cargo.toml
@@ -12,3 +12,4 @@ documentation.workspace = true
 anyhow.workspace = true
 axum = "0.8.1"
 sbv2_core = { version = "0.2.0-alpha6", path = "../sbv2_core" }
+tokio = { version = "1.44.1", features = ["full"] }
diff --git a/crates/sbv2_voicevox/src/main.rs b/crates/sbv2_voicevox/src/main.rs
index 8d6f05f..911d91b 100644
--- a/crates/sbv2_voicevox/src/main.rs
+++ b/crates/sbv2_voicevox/src/main.rs
@@ -1,5 +1,10 @@
+use axum::{routing::get, Router};
+use tokio::net::TcpListener;
 
+#[tokio::main]
 async fn main() -> anyhow::Result<()> {
-    println!("Hello, world!");
+    let app = Router::new().route("/", get(|| async { "Hello, world!" }));
+    let listener = TcpListener::bind("0.0.0.0:8080").await?;
+    axum::serve(listener, app).await?;
     Ok(())
 }