NVIDIA Video to Data
nvidia-isaac
A pipeline connecting human video reconstruction with simulation assets and robot-training data.
What you can explore
Inspect the reconstruction modules and physics-grounding workflow rather than assembling every stage independently.
Before you use it
Individual modules require their own dependencies and pretrained models. FoundationPose has separate model terms; output quality depends on reconstruction and robot configuration.
Original contributors
NVIDIA Video to Data contributors
License / access: Source code: Apache 2.0 · documentation/skills: CC BY 4.0 · pretrained models have separate terms
Editorial reference · checked 2026-09-22