feat: add more synthesize options

crates/yaejuyang-supertonic/src/main.rs

@@ -27,16 +27,28 @@ async fn main() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         .ok()
         .and_then(|v| v.parse().ok())
         .unwrap_or(1.05);
-    let silence_duration: f32 = std::env::var("SUPERTONIC_SILENCE_DUR")
+    let silence_before_duration: f32 = std::env::var("SUPERTONIC_SILENCE_BEFORE_DUR")
         .ok()
         .and_then(|v| v.parse().ok())
-        .unwrap_or(0.3);
+        .unwrap_or(0.0);
+    let silence_in_content_duration: f32 = std::env::var("SUPERTONIC_SILENCE_IN_CONTENT_DUR")
+        .ok()
+        .and_then(|v| v.parse().ok())
+        .unwrap_or(0.26);
+    let silence_after_duration: f32 = std::env::var("SUPERTONIC_SILENCE_AFTER_DUR")
+        .ok()
+        .and_then(|v| v.parse().ok())
+        .unwrap_or(0.4);
     let workers: usize = std::env::var("SUPERTONIC_WORKERS")
         .ok()
         .and_then(|v| v.parse().ok())
         .unwrap_or(2);
     let hf_repo = std::env::var("SUPERTONIC_HF_REPO")
         .unwrap_or_else(|_| "https://huggingface.co/Supertone/supertonic-3".to_string());
+    let sound_size_mul: f32 = std::env::var("SUPERTONIC_SIZE_MUL")
+        .ok()
+        .and_then(|v| v.parse().ok())
+        .unwrap_or(1.5);
 
     let model_path = PathBuf::from(&model_dir);
     tts::assets::ensure_assets(&model_path, &hf_repo)?;
@@ -66,7 +78,10 @@ async fn main() -> Result<(), Box<dyn std::error::Error + Send + Sync>> {
         TtsOpts {
             total_step,
             speed,
-            silence_duration,
+            silence_in_content_duration,
+            silence_before_duration,
+            silence_after_duration,
+            sound_size_mul,
         },
     )?);
 

crates/yaejuyang-supertonic/src/tts/engine.rs

@@ -189,7 +189,10 @@ impl TextToSpeech {
         style: &Style,
         total_step: usize,
         speed: f32,
-        silence_duration: f32,
+        silence_in_content_duration: f32,
+        silence_before_duration: f32,
+        silence_after_duration: f32,
+        sound_size_mul: f32,
     ) -> Result<(Vec<f32>, f32)> {
         let max_len = if lang == "ko" || lang == "ja" {
             120
@@ -198,8 +201,13 @@ impl TextToSpeech {
         };
         let chunks = chunk_text(text, Some(max_len));
 
-        let mut wav_cat: Vec<f32> = Vec::new();
+        let before_slience_len = (silence_before_duration * self.sample_rate as f32) as usize;
-        let mut dur_cat: f32 = 0.0;
+        let after_slience_len = (silence_after_duration * self.sample_rate as f32) as usize;
+        let content_silence_len = (silence_in_content_duration * self.sample_rate as f32) as usize;
+        let content_silence = vec![0.0f32; content_silence_len];
+        let after_slience = vec![0.0f32; after_slience_len];
+        let mut wav_cat: Vec<f32> = vec![0.0f32; before_slience_len];
+        let mut dur_cat: f32 = silence_before_duration + silence_after_duration;
 
         for (i, chunk) in chunks.iter().enumerate() {
             let (wav, duration) = self.infer(
@@ -216,14 +224,18 @@ impl TextToSpeech {
 
             if i == 0 {
                 wav_cat.extend_from_slice(wav_chunk);
-                dur_cat = dur;
+                dur_cat += dur;
             } else {
-                let silence_len = (silence_duration * self.sample_rate as f32) as usize;
+                wav_cat.extend_from_slice(&content_silence);
-                let silence = vec![0.0f32; silence_len];
-
-                wav_cat.extend_from_slice(&silence);
                 wav_cat.extend_from_slice(wav_chunk);
-                dur_cat += silence_duration + dur;
+                dur_cat += silence_in_content_duration + dur;
+            }
+        }
+        wav_cat.extend_from_slice(&after_slience);
+
+        if sound_size_mul != 1.0f32 {
+            for inner in wav_cat.iter_mut() {
+                *inner *= sound_size_mul;
             }
         }
 

crates/yaejuyang-supertonic/src/tts/pool.rs

@@ -13,7 +13,10 @@ use super::wav::wav_bytes;
 pub struct TtsOpts {
     pub total_step: usize,
     pub speed: f32,
-    pub silence_duration: f32,
+    pub silence_in_content_duration: f32,
+    pub silence_before_duration: f32,
+    pub silence_after_duration: f32,
+    pub sound_size_mul: f32,
 }
 
 pub struct TtsJob {
@@ -72,7 +75,10 @@ impl TtsPool {
                                     &style,
                                     opts.total_step,
                                     opts.speed,
-                                    opts.silence_duration,
+                                    opts.silence_in_content_duration,
+                                    opts.silence_before_duration,
+                                    opts.silence_after_duration,
+                                    opts.sound_size_mul,
                                 )
                                 .map_err(|e| e.to_string())?;
                             wav_bytes(&wav, tts.sample_rate).map_err(|e| e.to_string())