PAMI: part-anchored motion lifts contact recall 14.5% in text-to-HOI generation
_akhaliq · x · 2026-10-09
PAMI (Part Anchored Motion) generates human-object interaction motion from text by representing object motion relative to body-part anchors and learning to combine their predictions over time. A recursive Refiner operating in explicit 3D improves contact, grasping, and penetration. It improves contact recall by 14.5% over the previous state of the art, with videos and comparisons on the project page.
Related event: PAMI Boosts Text-to-HOI Motion Generation with Part-Anchored Objects(2 posts)→
More from Multimodal
- AI image of the day: GPT-6 Image 2.5 recreates 1940s Chinese village newsreel — DeryaTR_ · 2026-10-09
- Vigglorious Studio: chunked reference frames and guide keyframes fix character drift in long AI videos — Tablaski · 2026-10-09
- Star Wars AI songs strike again: Qui-Gon gets his own 'Gin and Juice' anthem — aronchick · 2026-10-09
- MiniMax H3 zombie short film goes viral on Reddit — gabxav · 2026-10-09
- Seedance 2.5 AI video praised as a "masterpiece" in viral share — BLUECOW009 · 2026-10-09
- HeyGen Video Tops OpenRouter's Video Rankings, Hands Out Codes for 100 Free Clips Each — aziz4ai · 2026-10-09