fix: voicevox化は難しいので、独自のエディター開発をする。

crates/sbv2_core/src/jtalk.rs

@@ -79,7 +79,7 @@ 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, Option<String>, String, i32)> {
+) -> Vec<(String, i32)> {
     let phones = &phones[1..];
     let tones = &tones[1..];
     let mut results = Vec::new();
@@ -90,7 +90,7 @@ fn phone_tone_to_kana(
         .zip(tones.iter().zip(tones.iter().skip(1)))
     {
         if PUNCTUATIONS.contains(&phone.clone().as_str()) {
-            results.push((phone.to_string(), None, "pau".to_string(), tone));
+            results.push((phone.to_string(), tone));
             continue;
         }
         if CONSONANTS.contains(&phone.clone()) {
@@ -99,8 +99,8 @@ fn phone_tone_to_kana(
             current_mora = phone.to_string()
         } else {
             current_mora += phone;
-            let (kana, consonant, vowel) = MORA_PHONEMES_TO_MORA_KATA.get(&current_mora).unwrap();
-            results.push((kana.to_string(), consonant.clone(), vowel.to_string(), tone));
+            let kana = MORA_PHONEMES_TO_MORA_KATA.get(&current_mora).unwrap();
+            results.push((kana.to_string(), tone));
             current_mora = String::new();
         }
     }
@@ -196,7 +196,7 @@ impl JTalkProcess {
         Ok((phones, tones, new_word2ph))
     }
 
-    pub fn g2kana_tone(&self) -> Result<Vec<(String, Option<String>, String, i32)>> {
+    pub fn g2kana_tone(&self) -> Result<Vec<(String, i32)>> {
         let (phones, tones, _) = self.g2p()?;
         Ok(phone_tone_to_kana(phones, tones))
     }

crates/sbv2_core/src/mora.rs

@@ -25,7 +25,7 @@ static MORA_LIST_ADDITIONAL: Lazy<Vec<Mora>> = Lazy::new(|| {
     data.additional
 });
 
-pub static MORA_PHONEMES_TO_MORA_KATA: Lazy<HashMap<String, (String, Option<String>, String)>> =
+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() {
@@ -35,11 +35,7 @@ pub static MORA_PHONEMES_TO_MORA_KATA: Lazy<HashMap<String, (String, Option<Stri
                     mora.consonant.clone().unwrap_or("".to_string()),
                     mora.vowel
                 ),
-                (
-                    mora.mora.clone(),
-                    mora.consonant.clone(),
-                    mora.vowel.clone(),
-                ),
+                mora.mora.clone(),
             );
         }
         map

crates/sbv2_editor/src/main.rs

@@ -1,4 +1,4 @@
-use axum::{extract::Query, routing::get, Router};
+use axum::{extract::Query, routing::get, Router, Json};
 use sbv2_core::{jtalk::JTalk, tts_util::preprocess_parse_text};
 use serde::Deserialize;
 use tokio::net::TcpListener;
@@ -12,11 +12,22 @@ struct RequestCreateAudioQuery {
     text: String,
 }
 
-async fn create_audio_query(Query(request): Query<RequestCreateAudioQuery>) -> AppResult<String> {
+#[derive(Deserialize)]
+struct ResponseCreateAudioQuery {
+    kana: String,
+    tone: i32,
+}
+
+async fn create_audio_query(Query(request): Query<RequestCreateAudioQuery>) -> AppResult<Json<Vec<ResponseCreateAudioQuery>>> {
     let (normalized_text, process) = preprocess_parse_text(&request.text, &JTalk::new()?)?;
     let kana_tone_list = process.g2kana_tone()?;
-    println!("{:?}", kana_tone_list);
-    Ok(normalized_text)
+    let response = kana_tone_list.iter().map(|(kana, tone)| {
+        ResponseCreateAudioQuery {
+            kana: kana.clone(),
+            tone: *tone,
+        }
+    }).collect::<Vec<_>>();
+    Ok(Json(response))
 }
 
 #[tokio::main]