Splat Maker

Scene

Viewing from

1Input

The photo and the views the engine uses when it creates new views. Click one to look from it.

2Place new frames

Move to a spot the scene does not cover well, then add a frame there. Up to 4 at a time.

3Create new views for frames

The engine looks from each frame, fills in what is missing, and adds it to the scene.

Drop a photo to start a scene

The photo becomes gaussians you can move around in. Depth is worked out in this browser (about 100 MB of model the first time); the parts out of sight are imagined on Cloudflare's servers, so the photo's views are sent there. It takes several minutes. Then you place frames and create new views to grow the scene.