deepgeek/CosyVoice
2
0
Fork 0
mirror of https://github.com/FunAudioLLM/CosyVoice.git synced 2026-02-04 17:39:25 +08:00
Code Issues Packages Projects Releases Wiki Activity

update tempo change

Browse Source
This commit is contained in:
lyuxiang.lx
2024-09-18 16:13:40 +08:00
parent c901a12789
commit e19e80fcd8
4 changed files with 23 additions and 31 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
cosyvoice/cli/cosyvoice.py
View File

@@ -53,43 +53,43 @@ class CosyVoice:
spks = list(self.frontend.spk2info.keys())
return spks
def inference_sft(self, tts_text, spk_id, stream=False):
def inference_sft(self, tts_text, spk_id, stream=False, speed=1.0):
for i in tqdm(self.frontend.text_normalize(tts_text, split=True)):
model_input = self.frontend.frontend_sft(i, spk_id)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.inference(**model_input, stream=stream):
for model_output in self.model.inference(**model_input, stream=stream, speed=speed):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
yield model_output
start_time = time.time()
def inference_zero_shot(self, tts_text, prompt_text, prompt_speech_16k, stream=False):
def inference_zero_shot(self, tts_text, prompt_text, prompt_speech_16k, stream=False, speed=1.0):
prompt_text = self.frontend.text_normalize(prompt_text, split=False)
for i in tqdm(self.frontend.text_normalize(tts_text, split=True)):
model_input = self.frontend.frontend_zero_shot(i, prompt_text, prompt_speech_16k)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.inference(**model_input, stream=stream):
for model_output in self.model.inference(**model_input, stream=stream, speed=speed):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
yield model_output
start_time = time.time()
def inference_cross_lingual(self, tts_text, prompt_speech_16k, stream=False):
def inference_cross_lingual(self, tts_text, prompt_speech_16k, stream=False, speed=1.0):
if self.frontend.instruct is True:
raise ValueError('{} do not support cross_lingual inference'.format(self.model_dir))
for i in tqdm(self.frontend.text_normalize(tts_text, split=True)):
model_input = self.frontend.frontend_cross_lingual(i, prompt_speech_16k)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.inference(**model_input, stream=stream):
for model_output in self.model.inference(**model_input, stream=stream, speed=speed):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
yield model_output
start_time = time.time()
def inference_instruct(self, tts_text, spk_id, instruct_text, stream=False):
def inference_instruct(self, tts_text, spk_id, instruct_text, stream=False, speed=1.0):
if self.frontend.instruct is False:
raise ValueError('{} do not support instruct inference'.format(self.model_dir))
instruct_text = self.frontend.text_normalize(instruct_text, split=False)
@@ -97,7 +97,7 @@ class CosyVoice:
model_input = self.frontend.frontend_instruct(i, spk_id, instruct_text)
start_time = time.time()
logging.info('synthesis text {}'.format(i))
for model_output in self.model.inference(**model_input, stream=stream):
for model_output in self.model.inference(**model_input, stream=stream, speed=speed):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
yield model_output