Task guides

Task overview

Choose the right Atlas task for what you have and what you want to make.

Start with the result you want. Each task links to its guide or API reference.

Task	When to use it
`atlasTextToImage`	Make a starting image from a written description of a place.
`images2PosedRGBD`	Estimate camera poses and depth for your images so you can use them to generate new views or build a 3D scene.
`atlasGenerate`	Explore the same scene from new camera angles. Start with one or more images that have a camera pose and a depth map.
`atlasChisel`	Give a rough 3D layout its appearance, restyle a scene, or edit selected parts of existing views.
`posedImages2Splats`	Build a 3D scene (a Gaussian splat) from images with camera poses and depth maps.
`splats2Mesh`	Convert a Gaussian splat into a GLB mesh, with optional texture or vertex colors and optional collider boxes.

Starting with a single photo? Run images2PosedRGBD first to get its camera pose and depth, then use atlasGenerate to create new views. The TypeScript quickstart walks through both steps.

For a finished video or 3D scene, the workflow guides chain the tasks: Build a scene from images, then Make a video or Make splats.

See the API reference for request fields and limits.