fix

crates/sbv2_core/src/tts.rs

@@ -41,7 +41,7 @@ pub struct TTSModelHolder {
     tokenizer: Tokenizer,
     bert: Session,
     models: Vec<TTSModel>,
-    jtalk: jtalk::JTalk,
+    pub jtalk: jtalk::JTalk,
     max_loaded_models: Option<usize>,
 }
 

crates/sbv2_editor/src/main.rs

@@ -35,9 +35,13 @@ struct ResponseCreateAudioQuery {
 }
 
 async fn create_audio_query(
+    State(state): State<AppState>,
     Query(request): Query<RequestCreateAudioQuery>,
 ) -> AppResult<impl IntoResponse> {
-    let (text, process) = preprocess_parse_text(&request.text, &JTalk::new()?)?;
+    let (text, process) = {
+        let mut 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()

test.py

@@ -0,0 +1,19 @@
+import requests
+
+
+data = (requests.get("http://localhost:8080/audio_query", params={
+    "text": "こんにちは",
+})).json()
+print(data)
+
+data = (requests.post("http://localhost:8080/synthesis", json={
+    "text": "こんにちは",
+    "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)