Chapter 4

Making the images

This chapter turns the previous three chapters into actual pictures: generating images per shot, comparing candidates, and adopting the best.

1. Generate per shot

In the generation step, select a shot and start an image generation. The system hands the image model that shot’s storyboard details, the confirmed character portraits, and the scene setup, so the result follows what you’ve already approved.

2. Candidates and adoption

Each run produces candidate images. A candidate is not a final — keep generating until one matches your vision, then adopt it as the shot’s official image. Unadopted candidates are kept, so you can always come back and swap.

3. Tips for better output

  • The more specific the shot card, the more controllable the image; fix vague shots in the storyboard first.
  • If a character drifts off-model, check that their portrait is confirmed before regenerating.
  • Image models differ in strengths and price — explore composition with a lighter model, then switch to a higher-quality one for the final take.

4. About video

Assembling images into finished video is rolling out gradually and isn’t covered here yet. Everything you’ve made — script, storyboard, characters, images — is kept intact and carries straight over once video is available.