Your video
What are we making?
One sentence is enough. The agent can inspect the footage and work out the details.
Planning
How should the agent start?
This controls whether you approve the direction before the agent builds the first edit.
Video building blocks
What can the edit use?
Pick several. These are presentation modes the agent can switch between, not rigid templates.
or choose below
your demoreal screen or screenshot
REAL SCREENLet the proof
take the frame.
take the frame.
MAKE THE POINT CLEAR
Agent decides which of these modes helps each moment.
Captions + sound
How much should the edit add?
Pick the intensity. “Auto” leaves the call to the agent.
CaptionsAgent decides how much captioning the video needs.
Sound designAgent decides whether sound effects help.
Dynamic captions move faster and emphasize selected words.
Clean captions stay simpler and calmer.
Lots of sound allows frequent clicks, impacts, pops, and whooshes, but speech still wins.
Transcription
How should the agent understand speech?
Transcription helps with cuts, timing, and captions. It is not mandatory for every video.
Never paste API keys into the prompt. The coding agent should use credentials stored locally and ask before sending private audio to a cloud provider.
Finish
When should the agent stop?
Preview-first is safer for creative work. One-shot is useful when you already trust the direction.