Refactor gen_audio function to use CosyVoice TTS instead of SFT

config.py

@@ -8,7 +8,7 @@ import torch
 
 # Audios
 os.environ['PYTHONPATH'] = f'{os.path.dirname(__file__)}/modules/CosyVoice/third_party/Matcha-TTS:{os.environ.get("PYTHONPATH", "")}' # add tabs/audios/modules/CosyVoice/third_party/Matcha-TTS to PYTHONPATH
-
+os.environ['PYTHONPATH'] = f'{os.path.dirname(__file__)}/modules/CosyVoice:{os.environ.get("PYTHONPATH", "")}' # add tabs/audios/modules/CosyVoice to PYTHONPATH
 
 css = """
 @import url('https://fonts.googleapis.com/css2?family=Poppins:wght@300;400;600&display=swap');

tabs/audios/events.py

@@ -155,7 +155,7 @@ def gen_audio(text, mode, sft_speaker = None, speaker_audio = None, voice_instru
         if not sft_speaker:
             raise gr.Error('Please select a speaker')
         
-        for i, j in enumerate(cv_base.inference_sft(
+        for i, j in enumerate(cv_sft.inference_sft(
             tts_text=text,
             spk_id=sft_speaker,
         )):