ByteDance Seed releases Lucida: turning indoor video into editable, simulation-ready 3D scenes
petewoodbridge · x · 2026-09-01
ByteDance Seed, with Peking University and Zhejiang University, released Lucida, a composable real-to-sim scene modeling system that reconstructs real indoor video into complete, editable object assets arranged as observed.
Pipeline:
- Parse: select keyframes, associate instances across views, and build a verified scene graph with masks, 3D boxes, and partial point clouds.
- Generate: synthesize an occlusion-free object image per instance and lift it via Seed3D 2.0 into a complete, editable 3D asset.
- Place: closed-loop GizmoAct refinement where a VLM operates a 3D editor, issuing executable pose edits until the rendered asset aligns with the capture.
It also provides VLM-based detection with referring descriptions, strong pose alignment, and high-fidelity real-to-sim digital twins where every object is a separate simulation-ready mesh.
Related event: ByteDance's Lucida Turns Indoor Video into Editable 3D Simulation Scenes(4 posts)→
More from Embodied
- SimInfinity: open-source robotic simulation benchmark with 10K real-world tasks — joemeno · 2026-10-18
- Robi: a German/English voice companion demo built on Gemma, Tyger and Humaneness Voice TTS — realmrfakename · 2026-10-18
- Munich's Plim unveils wheeled home-care humanoid controllable from 9,000 km away — CyberRobooo · 2026-10-18
- Musk says Optimus and Cybercab will make up the vast majority of Tesla's value — XFreeze · 2026-10-18
- SAE weight editing steers robot behavior at inference time, IIIT Hyderabad paper lands IJCAI Safe Physical AI workshop — ponguru · 2026-10-18
- NTU-led survey reviews nearly 200 papers on affordance learning for embodied AI — jiqizhixin · 2026-10-18