The world's first multi-modular world model: Lee Fei World Labs released Atlas, pixel-level precision control lens, and filmed the "bullet time"

SEPTEMBER 2ND, "AI GODMOTHER"Fei-Fei LiIdeas World Labs Today's launch, "The First World"MultimodalityworldModel"-- Atlas, CLAIMS THAT THE MODEL IS CAPABLE OF GENERATING IMAGES AND VIDEO FRAMES WITH PIXEL-LEVEL AND ACCURATE CAMERA CONTROL AND THAT IT IS REBUILT IN 3D。

The world's first multi-modular world model: Lee Fei World Labs released Atlas, pixel-level precision control lens, and filmed the "bullet time"

World Labs official presentation, where Atlas is pre-trained from scratch so that it can receive multi-modular input, including camera moves, and turn it into a 3D view。

The world's first multi-modular world model: Lee Fei World Labs released Atlas, pixel-level precision control lens, and filmed the "bullet time"

By positioning multiple input views in the spatial context of the model, Atlas allows users to generate images and video frames with precise controlGet a shot of a bullet time like that.

Atlas can also export a clear 3D model from one to more input images, going beyond the top open source reconstruction model。

Atlas will generate one or more reference images based on the location and angle of any camera you specifyI don't know. The resulting view corresponds to the content and geometry of the input image, allowing smooth extension of the image, and autonomously imagines the part of the scene that is not visible in the input。

Specifically, Atlas is able to perform a wide range of tasks covering world generation, reconstruction and simulation:

  • Camera control generation: Atlas generates images and videos from one or more images by pixel-precision camerasoutput 1440p video with maximum 1 minute.
  • Space reconstruction: Atlas can recreate the real scene from one to dozens of input images. It generates both a frame of images from an innovative perspective and a visible three-dimensional output, better than an advanced model dedicated to three-dimensional reconstruction。
  • Time and Space Simulation: Atlas redraws video to enhance dramatic visual effects by entering video modelling space and time and supports real-to-simulation workflows of robots。
  • Image Generation: Atlas supports the generation of images and 360-degree panorama from text, which follows complex tips, rendering text and generating visual styles。

World Labs officially indicated that some partners have acquired Atlas first-time eligibility and will open early access in the coming weeks。

1AI attached Atlas Official presentation address:

https://www.worldlabs.ai/blog/atlas

statement:The content of the source of public various media platforms, if the inclusion of the content violates your rights and interests, please contact the mailbox, this site will be the first time to deal with.
Information

Ali thousands asked Qwen3.8-Max-0902 release of the model to win the front-end programming champion

2026-9-2 13:25:38

Information

Meta moves internal communication from Google Chat to Slack, claiming that the latter is more appropriate for AI Agent

2026-9-3 8:36:51

Search