4DAnyone: Reconstruct a 4D human from a casual phone video, no rig needed
anselm · x · 2026-08-31
4DAnyone (Zhejiang University, Ant Group et al., SIGGRAPH Asia 2026) reconstructs a 4D Gaussian Splatting volumetric human from a single casually captured monocular video — no rig, no calibration, no tripod.
- Traditional photorealistic 4D human capture requires a calibrated synchronized camera array; this work instead generates the videos such an array would have recorded.
- Key challenge is consistency at scale: Reference Context Packing (RCP) compresses linearly growing reference context into a fixed budget, while Target Context Routing (TCR) routes context across disjoint generation groups to prevent structural drift; a 3D skeleton supplies sparse-but-accurate conditioning.
- Generalizes robustly to dance, sports, stage, speech and fashion footage, tolerating mild camera motion and unknown intrinsics/poses.
- Paper, code and models are released.
Related event: 4DAnyone: Reconstructing 4D Humans from a Single Phone Video(3 posts)→
More from Multimodal
- Indie Creator Builds an AI-Native Animated Show Solo — ComprehensiveLime654 · 2026-10-16
- Ideogram teases 4.5 arrival and open weights: 'Not yet, but soon' — East_Shoe8814 · 2026-10-16
- Opus 5.5 turns heads with surprisingly polished vertical video demo — FinanceYF5 · 2026-10-16
- 'Celestial Ocean Reverie' prompt template blends starry skies with dreamlike seascapes — LudovicCreator · 2026-10-16
- 'Thicc Lil' Red': an AI-made musical meme reimagining Little Red Riding Hood — bennash · 2026-10-16
- CoWorld Reasoner uses an evolving video world model as a visual reasoning engine — ModalityDance · 2026-10-16