Camera

A video has at most one camera, and it films every layer: text, shapes, images, video clips and custom components. Each layer sits at its own depth; when the camera moves, near layers move more than far ones (parallax). Move the camera for a push-in, orbit, pan or focus pull, not each layer.

Minimal push-in

Ask the agent, or call the tools through MCP (edit with op set to the tool name):

1. set_camera {} turns the camera on.
2. set_clip_depth { clip_id: "background", depth: 2500 } pushes the background far; { clip_id: "title", depth: -150 } brings the title forward. Depth is in pixels behind the focus plane, negative is closer.
3. apply_camera_preset { preset: "dolly-in", start: 0, duration: 3 } pushes in over three seconds.

At rest every layer keeps its size; during the push-in the title grows faster than the background.

set_clip_depth with space: "screen" keeps a caption or logo flat on top, ignoring the camera. Other presets: dolly-out, truck, pan, orbit, crane, dolly-zoom, rack-focus. set_camera_keyframes animates one camera value (x, y, z, rotateX/Y/Z, fov, focusDistance, aperture).

The canvas, the editor preview and exports draw the same camera move. Depth of field and lights show in the editor preview and exports only.