Handle speed

TTS/api.py

@@ -283,6 +283,7 @@ def tts(
             style_text=None,
             reference_speaker_name=None,
             split_sentences=split_sentences,
+            speed=1.0,
             **kwargs,
         )
         return wav
@@ -337,6 +338,7 @@ def tts_to_file(
             language=language,
             speaker_wav=speaker_wav,
             split_sentences=split_sentences,
+            speed=speed,
             **kwargs,
         )
         self.synthesizer.save_wav(wav=wav, path=file_path, pipe_out=pipe_out)

TTS/tts/models/xtts.py

@@ -379,7 +379,7 @@ def get_conditioning_latents(
 
         return gpt_cond_latents, speaker_embedding
 
-    def synthesize(self, text, config, speaker_wav, language, speaker_id=None, **kwargs):
+    def synthesize(self, text, config, speaker_wav, language, speaker_id=None, speed=1.0, **kwargs):
         """Synthesize speech with the given input text.
 
         Args:
@@ -436,6 +436,7 @@ def full_inference(
         gpt_cond_chunk_len=6,
         max_ref_len=10,
         sound_norm_refs=False,
+        speed=1.0,
         **hf_generate_kwargs,
     ):
         """
@@ -484,6 +485,7 @@ def full_inference(
             max_ref_length=max_ref_len,
             sound_norm_refs=sound_norm_refs,
         )
+        self.speed = speed
 
         return self.inference(
             text,
@@ -518,6 +520,7 @@ def inference(
         enable_text_splitting=False,
         **hf_generate_kwargs,
     ):
+        speed = self.speed
         language = language.split("-")[0]  # remove the country code
         length_scale = 1.0 / max(speed, 0.05)
         gpt_cond_latent = gpt_cond_latent.to(self.device)

TTS/utils/synthesizer.py

@@ -265,6 +265,7 @@ def tts(
         reference_wav=None,
         reference_speaker_name=None,
         split_sentences: bool = True,
+        speed=1.0,
         **kwargs,
     ) -> List[int]:
         """🐸 TTS magic. Run all the models and generate speech.
@@ -391,6 +392,7 @@ def tts(
                         d_vector=speaker_embedding,
                         speaker_wav=speaker_wav,
                         language=language_name,
+                        speed=speed,
                         **kwargs,
                     )
                 else: