Start a new Timeline-as-Data project.

Load a project from JSON.

Save the project to JSON.

Read the timeline at a zoom level (the agent inspects its work).

Compile the timeline to video (deterministic) and optionally QC it.

Run deterministic QC on a rendered file.

Ingest a media file: detect scenes, transcribe, analyze, index.

Build cross-asset narrative index from all ingested assets.

Search the semantic media index by text.

Remove silent portions from clips on a track.

Reframe a video asset for a different aspect ratio.

Snap clip transitions to the nearest beat in the audio track.

Generate captions from the transcript.

Automatically duck music volume when voice is present.

Assemble a cut from a structured edit plan (JSON string).

Create a short clip from source material matching a topic.

Run VLM semantic review on a rendered video.

Render a fast low-resolution proxy for review.

Run the multi-agent swarm to edit a video from a brief.

Export the project as an OpenTimelineIO (.otio) file.

Import an OpenTimelineIO file into the current project.

Render using the distributed render farm.

Show what the render farm would do without rendering.

Register a source asset.

Add a track.

Place a slice of an asset on a track (returns a diff).

Adjust a clip's source in/out.

Move a clip on its track.

Split a clip at timeline time t.

Delete a clip and close the gap.

Set clip playback speed.

Set clip volume.

Add a fade in/out to a clip edge.

Add a burned-in caption/title.

Add an audio clip (music bed / VO).

Snapshot current state.

Revert to the last checkpoint.