Video Composition

Workflow ownership

This owns the workflow after HyperFrames. Work from the brief and sources alone, and edit the starter-based scenes inline. For a revision, retain the accepted project, scene sources, palette, voice ID, complete content font, and pinned runtime. Update the scene contract and narration for the requested change; use the existing-project path in AUTHORING.md instead of bootstrapping again.

Fast production workflow

1. Lock the scene contract, then start media immediately

Record one semantic or narration beat per scene:

scene-id | Router layout id | presenter: off|talking-avatar | viewer outcome

Fix silent/static scene timing now. For voice or a talking avatar, open VOICE-AVATAR.md and generate media alongside draft assembly: authoring runs on provisional windows while media generates, then both join on measured seconds. Measured media replaces provisional estimates; it does not override a binding duration or preservation constraint. Dispatch media first once those constraints have a feasible plan.

2. Select from the executable layout library

Read only ROUTER.md with this file; it maps all 40 layouts to capacity, official items, motion, and presenter guidance.