feat: g2kana_tone

crates/sbv2_core/src/jtalk.rs

@@ -196,6 +196,11 @@ impl JTalkProcess {
         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 {

crates/sbv2_core/src/tts_util.rs

@@ -10,13 +10,12 @@ use tokenizers::Tokenizer;
 pub fn preprocess_parse_text(
     text: &str,
     jtalk: &jtalk::JTalk,
-) -> Result<(Vec<String>, Vec<i32>, Vec<i32>, String, JTalkProcess)> {
+) -> Result<(String, JTalkProcess)> {
     let text = jtalk.num2word(text)?;
     let normalized_text = norm::normalize_text(&text);
 
     let process = jtalk.process_text(&normalized_text)?;
-    let (phones, tones, word2ph) = process.g2p()?;
-    Ok((phones, tones, word2ph, normalized_text, process))
+    Ok((normalized_text, process))
 }
 
 /// Parse text and return the input for synthesize
@@ -35,8 +34,8 @@ 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, normalized_text, process) =
-        preprocess_parse_text(text, jtalk)?;
+    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);

crates/sbv2_voicevox/src/main.rs

@@ -13,9 +13,8 @@ struct RequestCreateAudioQuery {
 }
 
 async fn create_audio_query(Query(request): Query<RequestCreateAudioQuery>) -> AppResult<()> {
-    let (phones, tones, _, normalized_text, process) =
-        preprocess_parse_text(&request.text, &JTalk::new()?)?;
-    println!("{:?}", phones);
+    let (normalized_text, process) = preprocess_parse_text(&request.text, &JTalk::new()?)?;
+    let kana_tone_list = process.g2kana_tone()?;
     Ok(())
 }