Stanford's DyCraft Uses Agentic Decomposition to Cut Physical Errors in Long-Horizon 4D Scene Generation

StanfordUniversity · hf · 2026-11-11

Researchers from Stanford and Cambridge introduce DyCraft, an agentic system for long-horizon 4D scene generation.

• Core idea: decompose long-horizon 3D animation into planned static scenes, keyframe storyboards, and local motion authoring to reduce physical errors.
• Demo tasks include a robot assembling a flat-pack office chair in 13s, three movers carrying a piano upstairs in 112s, and a machine sorting six parcels by height with jam-handling.
• The staged agentic approach lets the system handle multi-step dynamic scenes that one-shot generation struggles with.

Original post →

More from Multimodal

• Fable 5.2 impresses with pixel art gaussian splats, posts go viral — Flomerboy · 2026-11-12
• Seedream 5.5 nails a consistent character sticker set at just $0.02 per image — azed_ai · 2026-11-12
• Ant Group's open-source Ming-3.0-tts-0.5B tops Open TTS Leaderboard with 0.88 WER — NielsRogge · 2026-11-12
• Viggle Remake recreated in ComfyUI: swap any character in your videos with MiniMax H3 — TheLocalLab · 2026-11-12
• Pika Integrates Seedream 5.5: Split Images Into 100+ Layers for Precise Multi-Turn Edits — aziz4ai · 2026-11-12
• Llama app adds simultaneous voice, image and video input, running locally on LFM3-Omni — mervenoyann · 2026-11-12

Multimodal channel →