Your device reads the depth, turns the frame into 3D and flies a real camera through it. Tap the picture to set focus.
Depth Anything V2 estimates depth on your device. Chucksi turns the still into a displaced mesh, cuts it at depth cliffs, fills a background layer behind foreground objects, and flies a virtual camera through it. Lens, film and grade are applied in a finishing pass, and WebCodecs encodes H.264 MP4 frame by frame.
Effects use the same depth. Chucksi finds the subject from what stands out toward the camera, fills in the scene behind it, and redraws the shot as a point cloud, drifting ash or trailing copies. No video model runs.
Limits: people in the frame can't act, and big moves stretch the background behind foreground objects. Keep intensity moderate for clean results.
Depth: Depth Anything V2 Small (Apache-2.0), in your browser. Camera, lens and film: Chucksi engine (MIT), in your browser. Encoding: your browser's H.264 encoder. Cost: $0.