SuperAPI AI Directory

cover

Design & creation

View repository

Updated 10-15 22:14

About this skill

Generate cover art prompts for your book using the cover-artist agent.

What This Does

  1. Asks whether you want image-only or text-inclusive covers
  2. Reads your project materials (README, themes, characters, world, tone)
  3. Synthesizes the story's visual essence
  4. Produces 3-5 distinct cover concepts as image generation prompts
  5. Optimizes for Kindle dimensions (1600 x 2560, aspect ratio 1:1.6)

Usage

/fiction:cover                    # Generate cover prompts for current project
/fiction:cover /path/to/project   # Generate for specific project

Text Options

You'll be asked whether to include text:

  • Image only (recommended) — Clean image with negative space for title overlay in post-production. More control over typography.
  • Include text — Title and author name generated as part of the image. Works well with GPT Image 1.5 and Gemini 3 Pro.

Output

For each concept, you'll receive:

  • Approach — The visual strategy
  • Prompt — Ready to paste into GPT Image, Gemini, FLUX, or other generators
  • Why it works — Genre signals and design rationale
  • Variations — Style/mood alternatives

Generating Images

Use the motif CLI to generate cover images directly:

motif "your prompt here" --cover          # Kindle cover: 2:3, 2K
motif "your prompt here" --cover -m gpt   # GPT model (best for covers)
motif "your prompt here" --cover --dry-run # Preview cost first

For illustrated books: After generating your cover, use /fiction:illustrate to create a motif series from the cover and generate consistent chapter illustrations in the same style.

Prompt Compatibility

Prompts are optimized for:

  • GPT Image 1.5 — Conversational prompts, specify dimensions
  • Gemini 3 Pro Image — Natural language, descriptive prompts
  • FLUX.2 (via fal.ai / motif) — Subject-first prompts, use --cover flag
  • Other generators — Adapt aspect ratio parameters as needed

Design Philosophy

The agent prioritizes:

  • Symbolic over literal — Evoke the feeling, don't illustrate the plot
  • Cut-through — Stand out in crowded genre thumbnails
  • Genre signals — Honor conventions while avoiding cliches
  • Thumbnail test — Readable at 80px width
  • Title space — Negative space in upper third for text overlay

After Generation

If you chose image-only:

  1. Generate images with your preferred AI tool (or falcon)
  2. Select the strongest result
  3. Add title and author name in a design tool (Canva, Photoshop, etc.)
  4. Test at thumbnail size
  5. Upload to KDP

If you chose text-inclusive:

  1. Generate images—inspect text carefully for errors
  2. Regenerate if any letters are wrong
  3. Minor touch-ups in design tool if needed
  4. Test at thumbnail size
  5. Upload to KDP

When to Use

  • Final stages of manuscript completion
  • Before submitting to beta readers (motivational)
  • Marketing preparation
  • Series branding exploration