Published by Video Runtime · Source check 2026-09-21 · Evidence method · Suggest a correction
EDITORIALLY VERIFIED PRODUCT INFORMATION
About World Labs Atlas
An omni world model for spatial intelligence with pixel-perfect camera-controlled image and video generation. Atlas is pretrained from scratch as a multimodal autoregressive diffusion transformer that combines text, images, video and 3D into a shared spatial context, then generates consistent new views and worlds from that context.
Key features
One model spans camera-controlled generation, spatial reconstruction with explicit 3D outputs, space-time simulation and text-to-image or 360-panorama generation. The announcement describes outputs of up to one minute of 1440p video from one or more reference images.
Getting started
Open the official project page and confirm current eligibility, region, account and payment requirements. Begin with the provider's own preview or onboarding flow; availability can change after this source check.
Real-time interaction
Atlas is pretrained from scratch as a multimodal autoregressive diffusion transformer that combines text, images, video and 3D into a shared spatial context, then generates consistent new views and worlds from that context.
Pricing
No public pricing; early-access requests only No public pricing; early-access requests only Verify the provider's current account, region, quota and billing terms before starting a paid or production session.
Limitations
There is no public product page pricing, no developer API and no hands-on access; the checked page offers only an early-access request. The checked page does not publish latency or FPS figures, so real-time suitability is unverified.
EDITORIAL VIEW
Editor's Verdict
World Labs Atlas is relevant to real-time generative video, interactive worlds or continuous visual agents, with access limits recorded from its official source.
Treat preview footage and provider benchmarks as evaluation leads, not proof of availability or performance for your workload.
How It Works
- Atlas is pretrained from scratch as a multimodal autoregressive diffusion transformer that combines text, images, video and 3D into a shared spatial context, then generates consistent new views and worlds from that context.
What We Like
- One model spans camera-controlled generation, spatial reconstruction with explicit 3D outputs, space-time simulation and text-to-image or 360-panorama generation.
- The announcement describes outputs of up to one minute of 1440p video from one or more reference images.
Current Limitations
- There is no public product page pricing, no developer API and no hands-on access; the checked page offers only an early-access request.
- The checked page does not publish latency or FPS figures, so real-time suitability is unverified.
Pricing & Access
- No public pricing; early-access requests only
- Verify the provider's current account, region, quota and billing terms before starting a paid or production session.
Verification Summary
- Official provider material reviewed on 2026-09-21. Video Runtime did not create an account, pay, or run a performance test.
- All speed, quality, scale and price statements are attributed to the provider and are not independent measurements.
Technical notes & official performance
Official Performance Data
- World Labs states Atlas outputs up to 1 minute of 1440p video with pixel-perfect camera control and improves with training compute. These are provider claims.
Technical Notes
- Atlas is announced as the model that will power future versions of World Labs' Marble product; the announcement is dated September 1, 2026.
Sources
- Official project or product page ↗official-site · accessed 2026-09-21
Access, stage & mechanism
These are separate source-checked facts, not a live uptime monitor. Unverified fields are left open; transport and persistent sessions do not establish how frames are generated.
- Release stage
- Not verified
- Access mode
- Not verified
- Source availability
- Not verified
- Transport
- Not verified
- Interaction
- Not verified
- Session mode
- Not verified
- Generation mechanism
- Not verified
- Continuity method
- Not verified
- Licence status
- Not verified
- Tasks
- Not classified
CONTINUE EXPLORING
Related projects & models
Selected by research links and category, not by popularity.
Create and navigate generated environments in real time through the experimental Project Genie experience.
A related project selected in this entry’s research.Experimental Project Genie access; account, product and regional conditions may apply
A public research preview of a narrow-domain playable world model that streams newly generated frames.
A related project selected in this entry’s research.Public research experience; account or capacity conditions may apply
Explore branching picture-world adventures where each choice creates a scene and a short film transition.
Another clip ahead option in this section.Public adventures; creating private worlds requires sign-in and scene packs
Create H3 Max video clips inside ElevenCreative, alongside ElevenLabs voice and audio tools.
Another clip ahead option in this section.Hosted generation; account and paid plan requiredRead how official-source and hands-on records stay separate in our Methodology.