Source: yang0/handraw-style on GitHub
yang0 got tired of image models drifting every time they asked for “a bit cuter, a bit more artsy” and numbered everything instead: hundreds of hand-drawn styles, layouts and single colors, each with a code. You pick a style, a layout and a color, add a topic, and the tool writes a bilingual prompt for Midjourney, DALL-E 3 or Flux. You can also describe what you want and let it choose. The README is mostly in Chinese, with an English version linked.
TL;DR
- 280 hand-drawn styles, 36 single colors (Klein blue, sage green, Hermes orange, Payne’s grey)
- 136 layouts: 21 social cards, 35 infographics, 68 comic storyboards, 12 IP design
- Output: Chinese and English image prompt for Midjourney, DALL-E 3, Stable Diffusion, Flux
- Install: give Codex the repo URL, ask it to install the skill
- Precise mode: name style, layout, color + topic (041, SC-001, C-01)
- Recommendation mode: describe the picture, skill picks style and color with a reason
- Article cover: ~200-word summary, visual metaphor; 2.35:1 or 21:9, 5:2 for X, 3:4 Xiaohongshu
- Article illustration: 2-5 images, one style and color, 4:3, can insert into Markdown
- Direct generation tuned for gpt-image-2; other models fall back to reference images
- Free to use, modify, sell; credit yang0 and repo URL; 4,001 stars, no detected license