World Labs Atlas is open to everyone: 3D worlds and video from a single photo
On Thursday, November 12, 2026 World Labs, the artificial intelligence company co-founded by Fei-Fei Li, opened Atlas, its world model, to everyone. Until that day only partners could use it. Now anyone can sign in, upload one photo or a set of photos of a place and get an explorable 3D world, camera paths and video clips. Here is what it does, how to use it and what we still do not know.
Leggi questo articolo in italiano
What World Labs Atlas is and what changed on November 12
World Labs Atlas is a world model: an artificial intelligence model that creates or reconstructs spatially consistent 3D environments you can move a camera through and turn into images, video and 3D scenes. World Labs introduced it on September 1, 2026 as an โomniโ model trained from scratch on text, images, video and 3D together.
In September access was limited to partners. On November 12, 2026 the company published the page โAtlas: Releasing our next generation world modelโ and opened it to anyone who signs up, with two web apps, Viewfinder and Chisel, plus the API endpoints behind them. Virtual production outlet VP Land summed it up this way: anyone can now build a 3D world from a single image, a set of location photos or a blocked-out layout.
The difference from a video generator such as Kling or Veo matters. A generator outputs a clip; Atlas first builds a 3D space, and the clip comes later, when you decide where the camera goes. That is why every shot of the same scene stays consistent with the others.
Viewfinder, Chisel and the other tools: what each one does
You use Atlas through four tools: Viewfinder for worlds from photos, Chisel for building a scene from blocks, Splats to Mesh for real geometry and Relight for lighting.
| Tool | What it does | Useful for |
|---|---|---|
| Viewfinder | Generates a world from a single image or reconstructs a real place from up to 64 overlapping photos; then you explore new viewpoints, create camera paths and render video | Video, location scouting, previz |
| Chisel | You block out the scene with planes, blocks and simple shapes (grayboxing), or start from a prompt or reference images; move objects and every camera view updates | Sets, architecture, design |
| Splats to Mesh | Converts splats into meshes with real geometry | Games, interactive environments, 3D software |
| Relight | Lets you change the lighting of splats | Look and mood of the scene |
According to World Labs' announcement as reported by VP Land, the outputs are keyframes, video clips and 3D Gaussian splat scenes, a technique that describes an environment as millions of small colored, semi-transparent points. The September technical post states that Atlas generates up to 1 minute of camera-controlled video at 1440p from one to six input images.
API, Blender and the real-time version
Every feature of the apps is also available as a public API, with a Blender integration: developers can build their own tools or point an AI agent straight at the model.
- Video formats. According to VP Land, the main API operation accepts target cameras at 1280ร720, 960ร720, 528ร704 and 720ร1280, so vertical formats for social media are covered too.
- Camera paths. The docs ask you to submit the full, ordered camera path in a single request.
- Real time. Since November 5, 2026 World Labs has shown Frameboy, a research preview that streams up to 15 frames per second to your browser while you steer the camera; the worlds are generated on cloud GPUs.
On the corporate side, on September 28, 2026 World Labs announced on its blog that it is joining AMD.
What it costs and the limits to know
As of November 14, 2026 World Labs has not published dedicated Atlas pricing, and the docs and API hostnames still carry beta labels. VP Land notes that the only reference point is the credit schedule of Marble, the company's earlier product, but Atlas may be priced differently.
- Stated numbers. In its September post World Labs states that human raters preferred Atlas over other video models between 75% and 94% of the time in camera-controlled comparisons. These are the company's own tests and have not been independently replicated.
- You need an account. Access goes through atlas.worldlabs.ai after signing up.
- Environments, not actors. The use cases World Labs lists (film and VFX, architecture and design, robotics) are about sets, spaces and environments; for people and performances it makes sense to pair Atlas with video generators and avatar tools.
What it means for you if you make video or content
For video makers, Atlas solves a concrete problem: keeping the same location consistent across several shots without regenerating the scene every time. World Labs pitches it directly for film and VFX: build the set, move the camera and render the shots you need.
- Scout with your phone. Take overlapping photos of a real place (up to 64) and rebuild it in Viewfinder: you get a consistent background to shoot from several angles.
- Use Atlas for the space, a generator for the action. Take keyframes of the location from Atlas and animate them with video models: read what changes with Kling 4.0 and Gemini Omni 1.2 Flash, and see our Kling AI guide.
- Try vertical formats. The API accepts 720ร1280: handy for Reels and TikTok.
- Compare world models. You can use Atlas today, Google's Genie 4 not yet: read our articles on Genie 4 and Higgsfield Playable.
The skill that matters is still directing: AI Filmaking teaches you to write scenes, shots and camera moves for AI video, the same ideas you will use to draw a camera path in Atlas. For quick results start with Create stunning AI videos in 5 minutes, and if you would rather have the video made for you there is the 30-second AI video service. For a complete method there is The AI Creator System, and here is how we use artificial intelligence.
Frequently asked questions
What is World Labs Atlas?
It is World Labs' world model, introduced on September 1, 2026 and opened to everyone on November 12, 2026: it creates or reconstructs 3D worlds from one or more images and turns them into frames, video and Gaussian splat scenes.
Can I use Atlas now?
Yes. Since November 12, 2026 anyone can sign up at atlas.worldlabs.ai and use the Viewfinder and Chisel apps; the same features are available as public API endpoints.
How much does Atlas cost?
World Labs has not published dedicated Atlas pricing. The docs and API are still labeled beta.
How many photos do I need to rebuild a place with Atlas?
One is enough to generate a whole world; to reconstruct a real place, Viewfinder accepts up to 64 overlapping photos.
What is the difference between Atlas and a video generator like Kling?
A video generator creates a clip. Atlas first builds a consistent 3D space and then renders video along the camera path you choose, so different shots of the same scene stay consistent with each other.
Read next



Sources checked on November 14, 2026
- World Labs โ Atlas: Releasing our next generation world model (12 novembre 2026)
- World Labs โ Atlas: A World Model for Spatial Intelligence (1 settembre 2026)
- World Labs โ Frameboy: Atlas in Real-time (5 novembre 2026)
- VP Land โ World Labs' Atlas World Model Opens to Everyone, Turning Photos Into Video and 3D Gaussian Splats (13 novembre 2026)