make-silhouette

Convert one named subject into a clean solid silhouette while preserving its recognizable outer contour and requested background or transparency. Use for people, animals, products, icons, or cut-paper

SKILL.md

GitHub

Make Silhouette

Catalog

  • Display name: Make Silhouette
  • Categories: Social Media, Portrait
  • Status: Ready
  • Route label: GPT Image 2 consistency-sensitive edit
  • Tone: slate
  • Short description: Turn a subject into a recognizable silhouette.

What this skill does

  • Isolates one target and converts its interior to a uniform fill.
  • Preserves gesture-defining contour details and optional negative spaces.

How to use

Provide source, target, fill color, background, and internal cutout rule.

Make a silhouette

Turn only the cyclist into solid black on the sunset; retain bicycle spoke gaps.

Workflow

  1. Define target contour, important negative spaces, ground contact, and background.
  2. Run gpt-image v1.0 with quality: "medium", imageSize: "2K", the supplied references, the requested or source-preserving ratio, batchCount: 1, and one result.
  3. Fill target uniformly; do not darken the whole image.
  4. Inspect edge continuity, accessories, negative spaces, halos, holes, and recognizable pose.
  5. Hand off to $remove-background when transparent alpha is requested.

User-facing output

  • Media type: One silhouette image
  • Default quantity: 1
  • Content per image: One target silhouette
  • Default layout: Source composition
  • Model policy: GPT Image 2 Medium/2K for consistency-sensitive final image editing
  • Downstream use: Icons, posters, and profile graphics