Skip to content

threerocks /hand-drawn-styles

Image prompts that keep one hand-drawn look across every episode
hand-drawn-styles screenshot from its README
From the hand-drawn-styles README on GitHub.

About hand-drawn-styles

It gives your AI agent 19 tested illustration recipes, like Ghibli, ink wash and crayon, and returns an image prompt rather than an image or video. Paste that prompt into gpt-image, Jimeng or Midjourney. For micro dramas, it keeps one drawn look across first frames and storyboards in every episode. The docs are mostly Chinese, and style 3.1 needs a reference image plus a three-step workflow.

Best for: Illustrated or comic-style micro dramas that need the same art style in every episode's key frames.

What it does

  • 19 recipes: 18 numbered styles plus a stable variant, 3.1
  • Pick a style by number, Chinese name or English alias like ghibli or pixel-art
  • Shows a style menu when you don't name one
  • Fills in subject, action and text slots from your plain description
  • scripts/render_prompt.py copies recipes word for word, so the look doesn't drift
  • Sample images for every style in the examples/ folder

Quickstart

Rewritten from the README. Check the repo for the latest steps.

  1. 1
    Send this line to your agent: the command below (it means "install this skill for me").
    帮我安装这个 Skill:https://github.com/threerocks/hand-drawn-styles
  2. 2
    Or for Claude Code by hand: the command below
    git clone https://github.com/threerocks/hand-drawn-styles.git ~/.claude/skills/hand-drawn
  3. 3
    For Cursor, Cline, Windsurf and others, copy PROTOCOL.md and STYLES.md into your rules or custom instructions.
  4. 4
    Ask for a picture in a named style, e.g. 用 ghibli 画…… or 用 3 号画风画……, then paste the prompt into your image model.
  5. 5
    For a fixed, unedited prompt: the command below
    python3 scripts/render_prompt.py --style 3.1 --subject '…' --text '不加任何文字' --aspect 3:4 --format json

Models and languages

Models it works with

  • gpt-image
  • Jimeng
  • Midjourney

Interface and docs

  • Chinese

Alternatives

Other agent skills projects people compare with hand-drawn-styles.

gnipbao

story-to-handdrawn-video

Agent skill that makes a hand-drawn 3:4 video from a story or images

Agent skills
  • OpenAI
2.1k GitHub stars+5/30d10 days agoMIT

smixs

visual-skills

Writes video and image prompts for Seedance, Kling and Veo

Agent skills
  • Seedance 2.5
  • Seedance 2.0
  • Kling 3.0 Turbo
  • +7
470 GitHub stars+8/30d16 days agoCC-BY-4.0

eternityspring

shuohao-skills

Pre-production assets for an AI short drama, straight from a novel

Agent skills
  • Qwen Image
  • GPT Image 2
  • MiniMax H3
  • +1
4.1k GitHub stars+34/30d6 days agoApache-2.0

zenstory-ai

drama-skills

Chains eleven agent skills into your whole AI short-drama pipeline

Agent skills
  • MiniMax H3
  • Seedance
  • GPT Image 2
  • +1
2.5k GitHub stars+31/30d4 days agoMIT

Featured on OpenMicroDrama

Do you maintain hand-drawn-styles? Add this badge to your README so English speakers can find our write-up.

Featured on OpenMicroDrama
markdown
[![Featured on OpenMicroDrama](https://openmicrodrama.com/badges/featured.svg)](https://openmicrodrama.com/projects/threerocks-hand-drawn-styles)

The badge links to this page. Want your description changed? Email support@openmicrodrama.com.

Reviewed Oct 3, 2026. We write these descriptions ourselves. The repo's own docs are the final word.