- 1
Define the story in beats
Ask AI Chat for a beginning, development and ending tied to the approved message. Give each beat one purpose. Establish what the viewer needs to understand without relying on a dense voiceover or unreadable generated text.
Checkpoint: Remove any scene that adds an unsupported product claim or does not support the core message.
- 2
Turn beats into a shot table
List subject, framing, action, camera direction, source asset and intended transition for each shot. Treat timings as editorial estimates. Put titles, captions and narration in separate columns so generation instructions remain focused on visual motion.
Checkpoint: A producer should know which assets already exist, which need permission and which are only proposed.
- 3
Adapt one shot to one route
Use the selected model guide to check prerequisites. For an image-to-video route, prepare a suitable starting image and describe a bounded motion relative to it. For another route, rewrite the brief according to its own input contract instead of copying provider-specific settings.
Checkpoint: Do not paste a full multi-scene screenplay into a single clip prompt and assume the route will edit it.
- 4
Review previews against the board
If you choose to generate or film footage, inspect each returned clip for action, continuity, framing and unwanted changes. Record actual duration and visible defects. Revise one property at a time, or choose a simpler shot rather than repeatedly submitting an unclear prompt.
Checkpoint: A generated preview is an observation to assess, not an automatic approval of fidelity, timing or usage rights.
- 5
Prepare the external edit handoff
Assemble approved clips, titles, narration, music and captions in your chosen editing workflow. Inspect the complete sequence for pacing and clarity. Keep a list of missing assets and an alternate shot for a concept that cannot be produced reliably.
Checkpoint: Check the finished export, including audio and captions, rather than approving only the individual source clips.