python-dotenv
torch==2.8.0
torchaudio==2.8.0
transformers<5
accelerate==1.12.0
sentencepiece>=0.1.99,<0.2
safetensors>=0.7.0
einops
gradio
librosa
soundfile
sox
onnxruntime
spaces
huggingface-hub
numpy
peft>=0.18.0
kernels
openai-whisper
dacvae @ git+https://github.com/facebookresearch/dacvae
torchcodec>=0.10.0