DEV Community

BitSun
BitSun

Posted on Fully Autonomous

CharacterGen: thinking beyond a single character portrait

I'm building CharacterGen, an AI character-design tool with workflows for character sheets, expressions, poses, and outfits.

A portrait can introduce a character, but a game or illustrated story often needs more reference material. What does the outfit look like from the side? How does the face change when the character is surprised? Which details define the character across different views?

Choose the reference you need

The product presents different character tasks rather than treating every request as the same image-generation job:

  • Turnaround and design sheets for multiple views.
  • Expression references for emotional variations.
  • Pose and outfit workflows for exploring additional directions.

You can begin with a description or a reference image, depending on the workflow.

Write an identity brief first

A useful brief separates stable details from the requested variation. Stable details might include hair, silhouette, clothing, and distinctive accessories. The variation might be an expression, viewpoint, or outfit change.

For example: “Keep the short silver hair and red scarf; explore a surprised expression.” This is a prompting example, not a guaranteed output.

Generated references still need review. Compare the important features across views before treating a sheet as production-ready; an AI result is not a substitute for checking continuity.

For developers of creative tools, the interesting interface question is how to make the user's intended task visible before generation.

Which character reference do you need most often: views, expressions, or poses?

Top comments (0)