Need pose references for AI image tools or figure practice, without a sketchy copyright pack? Shisei (姿勢 — “pose” in Japanese) is a free faceless white-mannequin pose library with multi-camera line art. The image set is CC0; the code is Apache-2.0. Online gallery shisei.indaloop.app, repo wquguru/shisei. Drop one camera angle into ChatGPT / ComfyUI / Midjourney (or similar) as a composition reference. Researched October 31, 2026.
At a glance
| Item | Details |
|---|---|
| Type | Free mannequin pose reference library (online gallery + downloadable zip) |
| Version | v0.3.0 when checked (2026-10-29) |
| License | Images / data CC0 1.0; code Apache-2.0 |
| Scale | About 280 poses × up to five cameras (front / three-quarter / side / top / bottom); QC-passed line art ~1,100+, plus some photo renders |
| Languages | Pose names and UI: Japanese / Chinese / English / Korean |
| Price | Free; GitHub Releases ships per-series and full zips (including w900 and original-size packs) |
| Stars | About 152 stars / ~45 forks when checked |
| Site / repo | shisei.indaloop.app · github.com/wquguru/shisei |
How to use and how series are split
Browse by series in the gallery: Full body · Standard, Upper body, and — blurred by default — Full body · Suggestive / Duo (author note: suggestive adult pose refs: faceless mannequin, no explicit detail; shown only after you open or toggle them; excluded from the preview carousel and search). Filter by angle, switch line art / photo, browse by pose or single frame; deep links open a specific camera view. Names and categories are in four languages for search.
Upload one angle as a pose / composition reference into your drawing tool; feeding several angles of the same pose usually improves character consistency. Face, clothes, and background still come from your prompt or other refs. The repo includes w300 / w600 / w900 web sizes; for larger files grab shisei-all or shisei-original from Releases. Every camera view is QC’d — fails are removed from the catalog with a reason, not a random dump.
This article treats Shisei as AI composition reference and figure-practice material, not an adult site; suggestive series stay blurred by default until you opt in. Content intent follows the author’s repo notes.
How it sits next to tools on this site
| Need | Better fit |
|---|---|
| Run open diffusion models locally | FLUX.2 Klein local setup |
| Squeeze more AI throughput from RTX 30 Ampere (Linux) | GPU Unlock |
| Pure Rust open-source image / PDF toolchain | storytold Craft suite (EN pending) |
| Clean, commercially friendly pose line-art refs | Shisei |
Who it is for (and caveats)
Good fit: AI art where you need a locked pose; learning to draw with multi-angle mannequins; want a clearly CC0 set you can zip offline.
Caveats: some series are marked suggestive and blurred by default — at work stick to Standard / Upper-body public sets; a /zh/ path 404s (use the home language switch or ?lang=zh); a few numbered poses are still being revised (roadmap); whether your generated images pass the target platform’s content policy is still on you.
Entry: Gallery · Releases downloads. Scale and licenses follow the October 31, 2026 README / LICENSE-images check.
