handraw-style: 365 numbered hand-drawn styles for AI image prompts

Describing a drawing style to an image model is harder than drawing it — the difference between a Parisian café sketch and a watercolor journal page evaporates in words. yang0’s handraw-style turns the whole vocabulary into numbers: styles 001–365 (watercolor, sketch, pop art, paper cut), 170 layout types (social cards, infographics, comic panels, IP design, e-commerce) and 36 classic monochrome theme colors, each with a sample image and tested bilingual prompts. Pick your numbers, add a subject, paste into Midjourney, Flux or any generator. The library grew from the 331 styles we recorded in our October 18 GitHub daily roundup to 365 within a month, and sat at 7,033 stars as of November 4, 2026 — about 1,300 stars added in thirty days.

Core features

• Three numbered dimensions: 365 styles, 170 layouts and 36 theme colors combine freely into prompts, so a whole content series can lock onto one visual system by pinning a set of numbers.
• Bilingual prompts: every combination outputs tested Chinese and English prompts, ready for either prompt language and any image tool.
• The repo is a Skill: SKILL.md plus STYLES.md, LAYOUTS.md and COLORS.md at the root let Claude Code and similar agents take “style 041, layout SC-001, autumn theme” as a direct instruction.
• Official Feishu knowledge base: a searchable gallery of all 365 style images, install troubleshooting, common generation errors and monetization case studies from creators.
• License plus one string: free to use, modify and commercialize; the only request is keeping the author name and repo URL — a standard-MIT-plus-attribution arrangement.

Typical use cases

• Series content at scale: lock a style/layout/color combination and every cover in a Xiaohongshu or newsletter series shares one visual identity.
• Client style pick: send the numbered gallery instead of iterating on “make it more artistic” feedback loops.
• Multi-panel comics: the comic-panel layout numbers keep visual consistency across an entire strip.

Quick start

git clone https://github.com/yang0/handraw-style.git

Open the gallery, choose style + layout + color numbers, then either hand the combination to an agent with the SKILL.md installed or paste the matching prompt into your image tool. Choice paralysis is the library’s one growing pain, and the ecosystem already answers it: lacemou/handraw-style-router (MIT) is a lightweight Router Skill that takes a subject and returns five differentiated style candidates, skipping the gallery crawl.

Summary

handraw-style fits creators who ship style-consistent social images, covers and comics daily; illustrators with their own visual language won’t need a numbered menu. Facts and limits: an HTML repo that works as-is, bilingual prompts tested against real generations; the MIT-plus-attribution license asks you to keep the author credit when selling; and at 331→365 styles in a month, expect the numbering to keep moving — cite the version you used. The hand-drawn prompt niche now has a small ecosystem around this repo, the Router being the first satellite; at roughly a thousand stars a month, it won’t be the last.