| # --- App (always needed) --- | |
| gradio>=5.49,<6 | |
| numpy>=1.26 | |
| pillow>=10.0 | |
| soundfile>=0.12 | |
| # --- Hugging Face ZeroGPU backend (live inference on a Space) --- | |
| spaces>=0.30 | |
| torch>=2.5 | |
| torchvision>=0.20 | |
| transformers>=5.4 | |
| accelerate>=1.0 | |
| qwen-vl-utils>=0.0.8 | |
| voxcpm>=0.1.0 | |
| misaki[zh]>=0.9 | |
| librosa>=0.10 | |
| sentencepiece | |
| protobuf | |
| # --- Optional alternative GPU backend --- | |
| modal>=1.0 | |