Subscribe
0%

Word Atlas: 7,298 illustrated IELTS words, Anki-grade spaced repetition, and an open dataset you can reuse

A polished, ad-free, offline-capable vocabulary app — and the more interesting part is that its words, illustrations and audio are released under CC BY 4.0 for anyone to build on.

Illustration for: Word Atlas: 7,298 illustrated IELTS words, Anki-grade spaced repetition, and an open dataset you can reuse
designfuchen/word-atlas

Illustrated IELTS vocabulary trainer with UK/US audio and FSRS spaced repetition; dataset released under CC BY 4.0.

Language
TypeScript
License
MIT (code), CC BY 4.0 (dataset)
Try it
git clone https://github.com/designfuchen/word-atlas.git && cd word-atlas && npm ci && npm run dev
View on GitHub

Most vocabulary apps are either ad-riddled or locked behind subscriptions. Word Atlas (词光), climbing GitHub's trending list this week, is neither — and it ships with something rarer: a large, permissively licensed dataset.

What's inside

  • 7,298 IELTS words, each with an original illustration, UK and US pronunciation, and two example sentences (14,596 sentences in total).
  • Four recall drills: pick the meaning, pick the word from an image, identify a word by ear, and spell it.
  • A typing studio for words, sentences and dictation, with WPM and accuracy tracked separately from memory stats.
  • A "word star map" — every learned word lights up a star, rendered with custom WebGL2 shaders (with a fallback for devices that can't).
  • Offline support as a PWA, shareable image cards, and JSON backup/restore.
  • No account, no ads, no tracking. Progress stays on your device.

The scheduling is the real engine

Word Atlas uses ts-fsrs, an implementation of FSRS — the same open-source scheduling algorithm used by newer versions of Anki. Instead of fixed review intervals, FSRS models how likely you are to forget each word and schedules a review just before that happens. You can adjust the target retention rate.

A nice detail: only your first answer to a card counts. Retrying until you get it right doesn't inflate your accuracy — which makes the stats honest.

The dataset is the bigger gift

The code is MIT-licensed, but the words, illustrations and audio are released under CC BY 4.0, with attribution to "Word Atlas contributors (词光)." The data ships as JSON and CSV in the data/ folder. That means teachers, app builders and researchers can reuse a large, illustrated, audio-backed English word list with just a credit line.

One caveat from the maintainers: the content was auto-validated but not manually reviewed, so expect occasional errors. Audio was synthesized with ElevenLabs.

Run it locally

git clone https://github.com/designfuchen/word-atlas.git
cd word-atlas
npm ci
npm run dev    # http://localhost:5188

Requires Node.js 20+. npm run build outputs a static site to dist/ you can host anywhere.

Our take

A genuinely well-crafted learning tool with a privacy-first design — and an open dataset that's likely to outlive the app itself. If you're studying for IELTS, use it. If you build education tools, look at the data/ folder.

Sources

  1. GitHub — designfuchen/word-atlas github.com
  2. GitHub — Latest 10 trending repositories, October 30, 2026 github.com

#open source#education#language learning#FSRS#TypeScript#open data