setuptools
huggingface_hub
bitsandbytes
transformers>=4.51.0
accelerate
sentence_transformers
diffusers>=0.30.0,<0.36.0
# Video export (required by diffusers export_to_video)
imageio[ffmpeg]
sentencepiece
openai
scipy
soundfile
# Local TTS
kokoro>=0.9.2
supertonic
# Local 3D generation (Shap-E ~300M params, ~6GB VRAM)
git+https://github.com/openai/shap-e.git
trimesh
# Note: Talking Head uses API only (SadTalker deps like dlib cause build timeout)