According to CoinMeta, World Labs founded by Li Feifei has released Atlas, which is officially claimed to be the world's first multimodal world model that can precisely control the camera to generate images and videos while also completing 3D reconstruction. Users only need to provide one to a few photos and draw the camera path, and Atlas will be able to fill in the areas behind walls, the back of objects, and the surrounding space that were not captured, generating videos up to 1 minute long at 1440p resolution. Unlike ordinary video models, Atlas directly reads the camera's position and angle in 3D space to create a deep and complete 3D scene. The more photos are input, the less content the model needs to supplement on its own. Atlas is currently only available to selected partners.