update space

app.py

@@ -76,7 +76,8 @@ tokenizer, model, image_processor, _ = load_pretrained_model(model_path, None)
 model = model.to('cuda').eval()
 model = model.bfloat16()
 
-tts_model = CosyVoice('iic/CosyVoice-300M-SFT', load_jit=True, fp16=True)
+# tts_model = CosyVoice('iic/CosyVoice-300M-SFT', load_jit=True, fp16=True)
+tts_model = CosyVoice('FunAudioLLM/CosyVoice-300M-SFT', load_jit=True, fp16=True)
 OUTPUT_SPEECH = False
 
 USE_SPEECH=False

ola/CosyVoice_main/cosyvoice/cli/cosyvoice.py

@@ -16,7 +16,8 @@ import time
 from typing import Generator
 from tqdm import tqdm
 from hyperpyyaml import load_hyperpyyaml
-from modelscope import snapshot_download
+# from modelscope import snapshot_download
+from huggingface_hub import snapshot_download
 import torch
 from cosyvoice.cli.frontend import CosyVoiceFrontEnd
 from cosyvoice.cli.model import CosyVoiceModel, CosyVoice2Model

requirements.txt

@@ -22,6 +22,7 @@ torchvision==0.16.1
 torchaudio
 modelscope
 matcha-tts
+pyworld
 av
 librosa
 gradio