HRuiCcc/RuiC-PoseAtlas

RuiC-PoseAtlas is a static reference library for creators who need to study a body pose from several camera angles at once. It organizes 288 pose sets into cards, each with front, three-quarter, side, top and low views, for AI-art prompting, 3D blockout, illustration and photography composition. The hosted demo is the fastest way to judge whether the content suits you.

What it offers

The library holds 1,439 images in total: 900 × 1200 PNG originals plus 300 × 400 JPEG thumbnails, grouped by pose number rather than scattered across categories. Each card shows the angles together, so proportions and limb direction can be compared without losing the pose identity. The grid loads thumbnails only; originals are requested when the lightbox opens or a download is triggered, and adjacent views are preloaded.

Search runs across Chinese, English and Japanese names, numbers, categories and series; filters cover four series, 26 subcategories and five angles. A single-image mode expands every angle for one-by-one review. The lightbox supports arrow-key navigation, Esc to close and a PNG download button, and deep links can point at a specific pose and angle. The interface language can be switched from a header control or pinned through ?lang=zh|en|ja, with the choice remembered in localStorage. See the interface preview and the live demo.

Running and deploying

The project is plain HTML, CSS and ES Modules with no framework, bundler or runtime dependencies. Browsing needs only a modern browser; for local use, serve the folder over HTTP rather than opening index.html directly, because the page fetches its pose data and file:// requests are blocked. Python 3 or any static server is sufficient. Node.js 18+ is needed only for regenerating data and running the verification script.

Deployment is a static copy of the site files. GitHub Pages is the current demo path and is automated on pushes; Cloudflare Pages is documented as an alternative (framework preset None, empty build command), and Netlify, Vercel, object storage or Nginx are also listed. For Cloudflare Pages, the documented limits are 25 MiB per file and 20,000 files per deployment, which the project's roughly 2,880 images stay below. If using GitHub Pages' branch mode, the docs advise exposing only site files so the README, scripts and donation image are not published.

A full clone is large: the PNG originals are about 654 MB and thumbnails about 25 MB, so expect a slow first pull. A fetch script can restore missing images and checks file signatures.

Licensing and content limits

The source code, styles, scripts and docs are MIT-licensed, but the images and original data are not. The original author permits creative use including commercial work, yet prohibits redistributing, reposting or reselling the image data itself. Anyone reusing the assets should read the acknowledgements and copyright section and the original site's terms. The project is an unofficial compilation, and the README notes its content is intended for adult creators.

The tool is a browsing reference, not a pose generator or 3D asset pipeline. The figures are faceless grey mannequin line drawings; faces, clothing and backgrounds are left to your own workflow. Customization means adding PNGs and records under scripts/source, adding English names, then running the data and verify scripts. The verification script checks counts, trilingual names, image dimensions, file references and i18n keys, but that is a repository-integrity check, not an independent quality or accuracy test.