A Chinese-language card catalog for AI drawing references
A translated, searchable front-end over three Japanese AI-art reference libraries — faces, poses, outfits — with a random-combo prompt generator bolted on.
What it does
This is a Chinese-localized web interface over three public reference libraries by Japanese creator @giaaibusi: 1,499 face portraits, 288 poses (1,439 images, five camera angles each), and 1,407 outfit references. Everything is translated and searchable in both Chinese and English, and you can copy ready-made drawing prompts with one click. The site itself is plain static JavaScript — serve it with any static file server and it runs.
The interesting bit
The random-combo feature is the clever part: it pulls one face, one outfit, and one pose, shows the three references side by side, and generates a prompt. It also handles the awkward split in today’s image models — agents that can fetch URLs get prompts with direct image links plus a curl command and an explicit “fetch first, don’t generate from description” instruction, while chat-only models get a single 2828×1380 triptych image (face / outfit / pose, numbered) you can paste directly.
Key highlights
- Three libraries behind one unified entry point; pose library supports depth variants
- Lock-and-reroll: swap just the face while keeping outfit and pose fixed
- Filter poses by category (standard, upper body, two-person) and outfits by season or style
- Bilingual Chinese/English search across all translated tags
- All reference images are AI-generated — no real people involved
Caveats
- Licensing is uneven: the face library is CC0, but the pose and outfit libraries forbid redistributing the images themselves (including modified versions) — this repo only ships thumbnails for search
- The outfit author did a brand-similarity check but doesn’t guarantee non-infringement; commercial use is your own judgment call
- The site runs GA4 and Microsoft Clarity analytics, if that matters to you
Verdict
Worth a bookmark if you prompt image models in Chinese and want structured references instead of vague adjectives. If you don’t read Chinese or already use the original Japanese sites, it’s mostly a translation layer — a genuinely useful one, but a translation layer.
Frequently asked
- What is sleepy-zone/ai-ref-library-cn?
- A translated, searchable front-end over three Japanese AI-art reference libraries — faces, poses, outfits — with a random-combo prompt generator bolted on.
- Is ai-ref-library-cn open source?
- Yes — sleepy-zone/ai-ref-library-cn is an open-source project tracked on heatdrop.
- What language is ai-ref-library-cn written in?
- sleepy-zone/ai-ref-library-cn is primarily written in JavaScript.
- How popular is ai-ref-library-cn?
- sleepy-zone/ai-ref-library-cn has 585 stars on GitHub.
- Where can I find ai-ref-library-cn?
- sleepy-zone/ai-ref-library-cn is on GitHub at https://github.com/sleepy-zone/ai-ref-library-cn.