world-labs-multi-image

To generate a coherent 3D world from multiple reference images, you provide up to four images with azimuth angles for explicit placement, or up to eight images for auto layout, alongside a descriptive world prompt.

Yes, you can control placement direction when generating 3D worlds by using the direction control mode, which allows you to specify azimuth angles for up to four reference images to ensure precise spatial positioning.

The maximum number of images for 3D space reconstruction is eight, available in auto layout mode where overlapping images are required to accurately reconstruct the desired architectural or spatial environment.

Use auto layout mode instead of direction control for 3D visualization when you have up to eight overlapping reference images to reconstruct spaces, whereas direction control is better for up to four images needing explicit azimuth placement.

Yes, multi-image world generation works with varied lighting conditions across references by supporting multiple inputs from the same space with consistent prompts that account for different lighting considerations during the 3D visualization process.