CMU's SNAP3D turns a single image into physically stable, snap-fit 3D printable parts
anselm · x · 2026-09-15
Researchers at Carnegie Mellon (Yu-Rou Tuan, Kris Kitani et al.) present SNAP3D, a physics-guided framework for part-aware 3D generation from a single image. Prior methods (XPart, OmniPart, PartCrafter) produce visually complete parts that interpenetrate and collapse under gravity. SNAP3D runs three stages: geometry repair to remove penetration, parameterized peg-and-socket connector generation against a recovered contact graph (the largest stability gain), and rigid-body physics simulation refinement. Assemblies are dropped into gravity simulation exactly as generated — baselines fall apart while SNAP3D holds. Feed it a photo and get glue-free, screw-free snap-fit parts ready for a 3D printer, demoed on robots, minifigures, furniture and more.
More from Multimodal
- Midjourney dual --sref combo produces medieval woodcut cat knight art — sergeantsref · 2026-11-02
- Fluffy AI-generated character designs: fox girl, alien bug and cat photographer — OVolosin82152 · 2026-11-02
- Fable 5.5 reportedly one-shots animation in pure code as Anthropic's coding bet pays off — Hesamation · 2026-11-02
- FlowHMR gets 5x faster, demoing an LLM agent pipeline from video to photoreal mocap — camenduru · 2026-11-02
- Retro 1973 music video "Rock is Dead" by TECHHALLA showcases startlingly realistic AI generation — Mother-Pear-2392 · 2026-11-02
- Weekend splatting with COLMAP and Lichtfeld — janusch_patas · 2026-11-02