Motion capture for any creature from one AI video: two cameras in one frame, a DINOv2 joint reader, the rig fitted in both views, a GLB out.
# Python 3.11. Versions this was built and tested with (October 2026).
# torch with CUDA comes from PyTorch's index: pip install torch==2.11.0 --index-url https://download.pytorch.org/whl/cu128
torch==2.11.0
transformers==5.20.2 # DINOv2 backbone (facebook/dinov2-with-registers-base, downloaded on first use)
numpy==2.4.6
opencv-python-headless==5.0.0.93
pillow==12.3.0
requests==2.34.2 # fal.ai queue client
matplotlib==3.11.2 # tools/plot_heights.py only