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 stream code

Browse Source
This commit is contained in:
lyuxiang.lx
2024-07-30 16:11:28 +08:00
parent 02f941d348
commit f4e70e222c
15 changed files with 182 additions and 109 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
cosyvoice/cli/cosyvoice.py
View File

@@ -49,6 +49,7 @@ class CosyVoice:
for i in 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):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
@@ -60,6 +61,7 @@ class CosyVoice:
for i in 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):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
@@ -72,6 +74,7 @@ class CosyVoice:
for i in 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):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))
@@ -85,6 +88,7 @@ class CosyVoice:
for i in self.frontend.text_normalize(tts_text, split=True):
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):
speech_len = model_output['tts_speech'].shape[1] / 22050
logging.info('yield speech len {}, rtf {}'.format(speech_len, (time.time() - start_time) / speech_len))