deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-05 01:49:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

update

Browse Source
This commit is contained in:
lyuxiang.lx
2024-12-12 16:46:28 +08:00
parent 2345ce6be2
commit c693039d14
6 changed files with 145 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
cosyvoice/cli/frontend.py
View File

@@ -142,11 +142,11 @@ class CosyVoiceFrontEnd:
model_input = {'text': tts_text_token, 'text_len': tts_text_token_len, 'llm_embedding': embedding, 'flow_embedding': embedding}
return model_input
def frontend_zero_shot(self, tts_text, prompt_text, prompt_speech_16k):
def frontend_zero_shot(self, tts_text, prompt_text, prompt_speech_16k, resample_rate):
tts_text_token, tts_text_token_len = self._extract_text_token(tts_text)
prompt_text_token, prompt_text_token_len = self._extract_text_token(prompt_text)
prompt_speech_22050 = torchaudio.transforms.Resample(orig_freq=16000, new_freq=22050)(prompt_speech_16k)
speech_feat, speech_feat_len = self._extract_speech_feat(prompt_speech_22050)
prompt_speech_resample = torchaudio.transforms.Resample(orig_freq=16000, new_freq=resample_rate)(prompt_speech_16k)
speech_feat, speech_feat_len = self._extract_speech_feat(prompt_speech_resample)
speech_token, speech_token_len = self._extract_speech_token(prompt_speech_16k)
embedding = self._extract_spk_embedding(prompt_speech_16k)
model_input = {'text': tts_text_token, 'text_len': tts_text_token_len,
@@ -157,8 +157,8 @@ class CosyVoiceFrontEnd:
'llm_embedding': embedding, 'flow_embedding': embedding}
return model_input
def frontend_cross_lingual(self, tts_text, prompt_speech_16k):
model_input = self.frontend_zero_shot(tts_text, '', prompt_speech_16k)
def frontend_cross_lingual(self, tts_text, prompt_speech_16k, resample_rate):
model_input = self.frontend_zero_shot(tts_text, '', prompt_speech_16k, resample_rate)
# in cross lingual mode, we remove prompt in llm
del model_input['prompt_text']
del model_input['prompt_text_len']
@@ -175,10 +175,10 @@ class CosyVoiceFrontEnd:
model_input['prompt_text_len'] = instruct_text_token_len
return model_input
def frontend_vc(self, source_speech_16k, prompt_speech_16k):
def frontend_vc(self, source_speech_16k, prompt_speech_16k, resample_rate):
prompt_speech_token, prompt_speech_token_len = self._extract_speech_token(prompt_speech_16k)
prompt_speech_22050 = torchaudio.transforms.Resample(orig_freq=16000, new_freq=22050)(prompt_speech_16k)
prompt_speech_feat, prompt_speech_feat_len = self._extract_speech_feat(prompt_speech_22050)
prompt_speech_resample = torchaudio.transforms.Resample(orig_freq=16000, new_freq=resample_rate)(prompt_speech_16k)
prompt_speech_feat, prompt_speech_feat_len = self._extract_speech_feat(prompt_speech_resample)
embedding = self._extract_spk_embedding(prompt_speech_16k)
source_speech_token, source_speech_token_len = self._extract_speech_token(source_speech_16k)
model_input = {'source_speech_token': source_speech_token, 'source_speech_token_len': source_speech_token_len,