NVIDIA researchers unveil JCA, co-generating 3D motion and video for arbitrary rigs
nvidia · hf · 2026-10-21
Researchers from NVIDIA, Technion, and Bar-Ilan University introduce JCA (Joints, Camera, Action!), a unified framework that co-generates 3D motion and video for arbitrary skeletons.
- Uses cross-attention with a pretrained video generator to couple 3D motion with video, producing scene-registered animation
- Supports bidirectional video-to-motion and motion-to-video conversion
- Handles multiple input modes: rig animation, keyframe-controlled first/last frame motion
The work bridges character animation, video generation, and motion capture, with direct implications for animation and embodied simulation workflows.
More from Multimodal
- UMELINA, Bratislava's first AI film festival, opens free submissions Nov 2026 — lmoroney · 2026-10-21
- AI agent scores Steve Jobs' Stanford speech in GarageBand fully via computer use — itsOmSarraf_ · 2026-10-21
- EasyMOSS workflow makes bbox region control trivial on MOSS in ComfyUI — Quiet_Quit_8894 · 2026-10-21
- Fish Audio raises $52M and launches open-weight emotive TTS model Drama 3 — fekdaoui · 2026-10-21
- Unlocking SA-Solver in ComfyUI: eta=1.7 beats Euler at 10 steps for low-step image generation — YakovSava · 2026-10-21
- Claude Code renders lyric videos frame-by-frame in Chromium, no AI video model needed — aitrendz_xyz · 2026-10-21