Tagbooru Toolkit

An all-in-one tagging and prompt formatting tool for anime generative AI workflows. Runs entirely in the browser — no backend required.

Live: https://xymoh.github.io/Tagbooru/

Features

Text Formatter

Paste noisy, unformatted text from any source and get clean, categorized Danbooru tag prompts.

• Parses messy input: line-based, mixed separators, trailing counts (29k, 7.6M, 665) stripped automatically
• Matches probable Danbooru tags via the Danbooru API and a local offline index of 1M+ tags
• Categorizes matched tags into organized prompt blocks:
  • Character
  • Series / Franchise (Source Material)
  • Copyright
  • Artist
  • General
  • Style / Quality
  • Looks / Appearance
  • Landscape / Scene
  • Action
  • Composition
  • NSFW
  • Meta
• One-click copy per category or copy all at once
• Source Material toggle to include/exclude franchise labels

Image Tagger (Standalone Desktop App)

A separate downloadable Windows application for local AI-powered image captioning. Available at img-tagboru-ai.

• Powered by WD14/SwinV2 models running locally on your GPU
• Batch processing of entire image folders
• Exports .txt caption files alongside images (LoRA training ready)
• Configurable thresholds, blacklists, and whitelists
• 100% offline — images never leave your machine

Privacy

• Text Formatter: Uses a pre-downloaded CSV index for offline tag lookup. Online Danbooru API calls are made only for tag matching (no images or personal data sent).
• Image Tagger: Runs entirely on your local machine via the standalone desktop app.

Tech Stack

• React + Vite
• No backend — static site hosted on GitHub Pages
• Danbooru API integration for tag matching
• SVG-based animated UI interactions

Development

npm install
npm run dev

Build

npm run build

Output goes to dist/.

Notes

• Danbooru API rate limits may apply with rapid successive lookups.
• Tag categorization is heuristic-based and continuously improved.
• The Image Tagger desktop app is a separate project: img-tagboru-ai.