Loading…
Please wait a moment.
Loading…
Please wait a moment.
Which inputs each system accepts, what it outputs, and whether you can run it in Roamscape today.
Models available in Studio, Compare, and Live today.
Multimodal world generation for persistent, explorable 3D scenes.
Physically grounded 3D world generation from text or image.
Next-gen real-time world model: minute-scale horizons, sub-second latency, camera + action control.
Real-time streaming video editing with a hybrid diffusion transformer, 24 FPS — try it live in your browser.
Real-time world model sessions with live video streaming and prompt steering.
Image-anchored navigable real-time world model with WASD-style control.
Real-time multi-shot video: one session, many scenes, soft shots and hard cuts.
Status, inputs, outputs, and access across the full index.
| Model | Status | Category | Inputs | Outputs | Access |
|---|---|---|---|---|---|
| MarbleWorld Labs | Live in Roamscape | Generative 3D world models | text · image · panorama · multi-view | navigable 3D world · SPZ · mesh · preview | Runnable in Roamscape |
| Echo-2spAItial | Live in Roamscape | Generative 3D world models | text · image · panorama | 3DGS scene · mesh · point cloud · semantic masks | Runnable in Roamscape |
| Genie 3Google DeepMind | Research | Interactive simulation world models | text · image preview / world sketch | interactive generated environment · real-time pixels | Limited access |
| LingBot-World-2Reactor | Live in Roamscape | Interactive simulation world models | reference image · text prompt · control signals (camera / action) | real-time navigable video stream (1664×960 @ 48 fps) | Runnable in Roamscape |
| SANANVIDIA · Reactor | Live in Roamscape | Interactive simulation world models | source video stream · text edit prompt · local edit regions · background replacement prompts | real-time edited video stream · 1280×704 streaming output · temporally consistent V2V frames · physical-AI augmented sensor video | Runnable in Roamscape |
| DreamX-WorldAMAP (Alibaba) | Research | Interactive simulation world models | image · text prompt · event prompts · camera control | generated interactive video / world | Open weights / code |
| HeliosReactor | Live in Roamscape | Interactive simulation world models | prompt · live commands | real-time video stream | Runnable in Roamscape |
| LingBotReactor | Live in Roamscape | Interactive simulation world models | reference image · text prompt · movement commands | real-time navigable video stream | Runnable in Roamscape |
| OdysseyOdyssey | Tracked | Interactive simulation world models | text · image · starting frame | interactive simulation · real-time pixels · world state | API available |
| Oasis 3Decart | Tracked | Physical AI / embodied world models | text prompt (scene) · throttle + steering actions | multi-camera RGB frames (left / front / right) | API available |
| LongLive-2.0Reactor | Live in Roamscape | Interactive simulation world models | text prompt (no reference image) | real-time multi-shot video stream (1280×704 @ 24 fps) | API available |
| DecartDecart | Tracked | Interactive simulation world models | video · text · interactive stream | real-time transformed video · interactive world simulation | Limited access |
| CosmosNVIDIA | Research | Physical AI / embodied world models | text · image · video · sound | video · synthetic data · action trajectories · world states | Open weights / code |
| HunyuanWorldTencent Hunyuan | Research | Open research systems | image · camera trajectory · text prompt | RGB-D video · depth · point cloud · 3D sequence | Open weights / code |
| V-JEPA 2Meta | Research | Physical AI / embodied world models | video · image goals · robot observations | latent predictions · planning signals · robot action support | Paper / research |
General-purpose real-time interactive world model.
Next-gen real-time world model: minute-scale horizons, sub-second latency, camera + action control.
Real-time streaming video editing with a hybrid diffusion transformer, 24 FPS — try it live in your browser.
General-purpose interactive world model with promptable world events and long-horizon memory.
Real-time world model sessions with live video streaming and prompt steering.
Image-anchored navigable real-time world model with WASD-style control.
Interactive world simulation from images and prompts.
Real-time multi-shot video: one session, many scenes, soft shots and hard cuts.
Real-time video/world models for low-latency immersive systems.
Real-time promptable driving world model — actions in, camera frames out — built for RL.
World foundation models for physical AI, robotics, and synthetic data.
Video-based world model for physical understanding and planning.
World models now span generative 3D, real-time simulators, and robotics pipelines. Roamscape tracks what each accepts, produces, and whether you can run it here.