gradio>=4.0.0
torch>=2.0.0
torchaudio>=2.0.0
transformers<4.50
transformers_stream_generator
soundfile
numpy<=1.26.4
vocos
torchcodec
cached_path
accelerate>=0.33.0
safetensors
tqdm>=4.65.0
matplotlib
pydub
huggingface_hub
x_transformers>=1.31.14
ema_pytorch>=0.5.2
torchdiffeq
librosa
jieba
pypinyin
datasets
click
tomli
hydra-core>=1.3.0
wandb