ByteDance research pipeline that turns indoor captures into editable, simulation-ready 3D scenes of separate object meshes.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Lucida is a research project from ByteDance for real-to-sim scene modeling. It takes a cluttered real indoor capture and reconstructs the scene as a set of complete, separately editable object assets arranged as observed, intended for use in simulation.
The pipeline has three stages: Parse, Generate and Place. Parse selects keyframes, associates object instances across views and builds a verified scene graph with reference views, masks, partial point clouds and 3D boxes. Generate uses each object's evidence to synthesize an occlusion-free image and lift it into a complete 3D mesh. Place initializes each asset coarsely and then refines its pose with GizmoAct, a vision-language model that operates a 3D editor in a closed loop, issuing one executable pose edit per turn until the rendered asset matches the captured scene.
The project page reports improvements in scene-level 3D object detection, object pose estimation and complete scene reconstruction on R2S, CA-1M and Aria Digital Twin. As of the provided material, only the project page and a technical paper are available; no code, weights, demo or pricing have been published.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Turns images of a messy room into an editable 3D simulation by identifying objects, generating complete meshes for each and placing them back. Only a technical paper is released so far.
Open-source background remover that handles glass, transparency, text, shadows and hair better than competitors shown. MIT license, under 900MB, with a free Hugging Face demo that took about 6 seconds.
Closest neighbours by meaning, not just by category.
Model · Nvidia
NVIDIA research model that turns a single image into a queryable 3D Gaussian scene with semantics and object instances.
Model · ByteDance
ByteDance's open-weights video generation and editing model pairing a Qwen2.5-VL planner with a Wan2.2 renderer.
Model
Feed-forward system that turns an RGB image or video into editable, textured per-object 3D meshes in under a minute.
Research · ByteDance
ByteDance research framework that distills multiple specialist image models into one via on-policy velocity matching.
Research
Research system that turns a single image into a 360° Gaussian scaffold and renders interactive video at 8 FPS.
Research
Research method that reconstructs PBR-ready indoor 3D scene meshes from multi-view images using a Trellis.2 generative prior.
