A static site that turns AI try-on into a browser-side affair
PoseWear bundles pose, outfit, and portrait reference libraries with bilingual prompts, then lets you generate try-on images in the browser using your own API keys.

What it does
PoseWear is a fully static reference library — pose line art, depth maps, realistic and anime outfits, and CC0 portrait references — each image tagged with Chinese and English prompts you can copy or combine. Pick an outfit and a pose, take them to the try-on studio, upload your own photo, and generate try-on images via OpenAI, Gemini, or Doubao Seedream. There’s no backend database and no sign-up: keys, favorites, and generation history live in your browser’s localStorage and IndexedDB.
The interesting bit
The repo ships zero images. A fetch script pulls gigabytes of assets from upstream galleries at build time, so the MIT license covers only code and copy — a clean way to sidestep licensing while still aggregating the material. The studio also rewrites your prompt into neutral e-commerce try-on language before sending, purely to reduce false positives in provider safety filters (it won’t bypass real moderation).
Key highlights
- Multi-provider support: OpenAI
gpt-image-1, Gemini’s Nano Banana, Doubao Seedream, and any OpenAI-compatible custom endpoint - 1320 anime outfits plus realistic outfits, poses, and depth maps, all searchable in Chinese and English
- PWA with offline caching of viewed content, dark mode, and bilingual UI
- Playwright test suite that mocks every provider’s failure modes — rate limits, timeouts, safety blocks — without real keys
- A Cloudflare Pages Function proxy exists only for CORS; by default the browser talks to providers directly
Caveats
- Building it yourself means downloading several GB of assets first; the repo alone won’t run
- Most outfit and pose images belong to upstream authors with no stated open license — commercial reuse requires contacting them
- The proxy endpoint is unauthenticated by design; the README suggests adding your own rate-limit rule in Cloudflare
Verdict
Worth a look if you do AI fashion or character generation and want a curated, prompt-ready reference shelf. If you were hoping for a self-contained image dataset you can redistribute, this isn’t that — deliberately so.
Frequently asked
- What is atuizz/posewear?
- PoseWear bundles pose, outfit, and portrait reference libraries with bilingual prompts, then lets you generate try-on images in the browser using your own API keys.
- Is posewear open source?
- Yes — atuizz/posewear is open source, released under the MIT license.
- What language is posewear written in?
- atuizz/posewear is primarily written in Python.
- How popular is posewear?
- atuizz/posewear has 1.4k stars on GitHub.
- Where can I find posewear?
- atuizz/posewear is on GitHub at https://github.com/atuizz/posewear.