The workflow
- Read the script aloud and cut anything you stumble over
- Check every fact against a source
- Split into scenes of four to ten seconds
- Write a one-line picture description for each scene
- Pick one visual style for the whole video
- Generate pictures for two or three scenes first and review the look
- Generate the rest, then voice, then captions
- Watch the full draft with sound off, then on
Why test two or three scenes first
A style mismatch discovered after generating forty images wastes credits. Early samples settle palette, line and character before volume.
What to fix in the script, not the picture
- Abstract sentences that have no obvious image
- Long sentences that run past ten seconds
- Names the voice may mispronounce