deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-05 18:09:24 +08:00
Code Issues Packages Projects Releases Wiki Activity

use wav file rather than tensor

Browse Source
This commit is contained in:
lyuxiang.lx
2025-12-08 08:43:09 +00:00
parent d985100326
commit 622a3a19b0
3 changed files with 31 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
cosyvoice/cli/cosyvoice.py
View File

@@ -67,9 +67,9 @@ class CosyVoice:
spks = list(self.frontend.spk2info.keys())
return spks
def add_zero_shot_spk(self, prompt_text, prompt_speech_16k, zero_shot_spk_id):
def add_zero_shot_spk(self, prompt_text, prompt_wav, zero_shot_spk_id):
assert zero_shot_spk_id != '', 'do not use empty zero_shot_spk_id'
model_input = self.frontend.frontend_zero_shot('', prompt_text, prompt_speech_16k, self.sample_rate, '')
model_input = self.frontend.frontend_zero_shot('', prompt_text, prompt_wav, self.sample_rate, '')
del model_input['text']
del model_input['text_len']
self.frontend.spk2info[zero_shot_spk_id] = model_input
@@ -89,12 +89,12 @@ class CosyVoice:
yield model_output
start_time = time.time()
def inference_zero_shot(self, tts_text, prompt_text, prompt_speech_16k, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
def inference_zero_shot(self, tts_text, prompt_text, prompt_wav, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
prompt_text = self.frontend.text_normalize(prompt_text, split=False, text_frontend=text_frontend)
for i in tqdm(self.frontend.text_normalize(tts_text, split=True, text_frontend=text_frontend)):
if (not isinstance(i, Generator)) and len(i) < 0.5 * len(prompt_text):
logging.warning('synthesis text {} too short than prompt text {}, this may lead to bad performance'.format(i, prompt_text))
model_input = self.frontend.frontend_zero_shot(i, prompt_text, prompt_speech_16k, self.sample_rate, zero_shot_spk_id)
model_input = self.frontend.frontend_zero_shot(i, prompt_text, prompt_wav, self.sample_rate, zero_shot_spk_id)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.tts(**model_input, stream=stream, speed=speed):
@@ -103,9 +103,9 @@ class CosyVoice:
yield model_output
start_time = time.time()
def inference_cross_lingual(self, tts_text, prompt_speech_16k, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
def inference_cross_lingual(self, tts_text, prompt_wav, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
for i in tqdm(self.frontend.text_normalize(tts_text, split=True, text_frontend=text_frontend)):
model_input = self.frontend.frontend_cross_lingual(i, prompt_speech_16k, self.sample_rate, zero_shot_spk_id)
model_input = self.frontend.frontend_cross_lingual(i, prompt_wav, self.sample_rate, zero_shot_spk_id)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.tts(**model_input, stream=stream, speed=speed):
@@ -129,8 +129,8 @@ class CosyVoice:
yield model_output
start_time = time.time()
def inference_vc(self, source_speech_16k, prompt_speech_16k, stream=False, speed=1.0):
model_input = self.frontend.frontend_vc(source_speech_16k, prompt_speech_16k, self.sample_rate)
def inference_vc(self, source_wav, prompt_wav, stream=False, speed=1.0):
model_input = self.frontend.frontend_vc(source_wav, prompt_wav, self.sample_rate)
start_time = time.time()
for model_output in self.model.tts(**model_input, stream=stream, speed=speed):
speech_len = model_output['tts_speech'].shape[1] / self.sample_rate
@@ -181,10 +181,10 @@ class CosyVoice2(CosyVoice):
def inference_instruct(self, *args, **kwargs):
raise NotImplementedError('inference_instruct is not implemented for CosyVoice2!')
def inference_instruct2(self, tts_text, instruct_text, prompt_speech_16k, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
def inference_instruct2(self, tts_text, instruct_text, prompt_wav, zero_shot_spk_id='', stream=False, speed=1.0, text_frontend=True):
assert isinstance(self.model, CosyVoice2Model), 'inference_instruct2 is only implemented for CosyVoice2!'
for i in tqdm(self.frontend.text_normalize(tts_text, split=True, text_frontend=text_frontend)):
model_input = self.frontend.frontend_instruct2(i, instruct_text, prompt_speech_16k, self.sample_rate, zero_shot_spk_id)
model_input = self.frontend.frontend_instruct2(i, instruct_text, prompt_wav, self.sample_rate, zero_shot_spk_id)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.tts(**model_input, stream=stream, speed=speed):