SAM 3D Body estimates a person's pose and body shape from a single photograph. It can give a character artist a starting point for studying proportions and reconstructing a pose, including hands and feet, through Momentum Human Rig. Optional 2D keypoint and mask prompts let you guide the estimate. The production work is still yours: a recovered body does not automatically reproduce the costume, facial likeness, textures or animation setup a game character needs. Use authorized adult reference photographs, check the recovered shape against your design, and build or adapt a conventional character asset around it.
Commercial use & license
SAM 3D Body code and checkpoints use the SAM License; redistribution and trade-control conditions apply. MHR is Apache-2.0. Audit exact detection, segmentation and field-of-view helpers separately; confirm terms for the default ViTDet weights before commercial use.
Adult content rules
The SAM License requires lawful use, privacy compliance and trade-control compliance. You still need permission for photographs, likenesses and third-party assets, and must check the exact helper checkpoints. Hosted adult-content rules are unverified; this license alone does not clear an adult-game project.
What you need
Python 3.11, PyTorch and the documented Detectron2 setup; approved, authenticated Hugging Face access to a SAM 3D Body DINOv3 or ViT-H checkpoint and its MHR asset. The default demo also uses MoGe2. Use authorized adult photographs and separately audited helpers. Account approval and hardware compatibility are untested.
Ranks are recommendations for the task described, based on reviewed documentation. Model ranks cover non-explicit development work. We have not run practical benchmarks. Check content policies and commercial terms separately. How the ranks work →
Sources & further reading +
- SAM 3D Body official repository ↗
- SAM 3D Body exact license ↗
- ViTDet helper checkpoint documentation ↗
Last checked 5 October 2026.