About this thing

A community-maintained dictionary of 1880 project, product, and programmer-jargon pronunciations; 1260 have cited sources.

Why

Every developer has been corrected mid-sentence at least once. "It's pronounced engine-x, not n-jinx." "It's koob-control, not kub-cuttle." "It's jif, not gif. The creator said so."

IPA helps you read the answer; audio helps you internalize it. We built this so you don't have to scrub through a 47-minute conference talk to find the right reading.

How the dictionary works

The whole thing is one TSV file: data/pronunciations.tsv. Ten columns per entry:

word | ipa | respelling_us | alt_ipa | alt_respelling_us
  | source_url | source_label | category | confidence | notes

Each entry is tagged with one of three confidence levels:

• creator-clarified — the source URL documents a creator quote or official FAQ entry. Best.
• community-consensus — widely used in the community, but no single citable source.
• contested — multiple readings genuinely compete. `say-it --all <word>` plays all of them.

Where the audio comes from

Every entry has a committed canonical .mp3 generated via macOS's built-in say command (say -v Samantha -r 175, with inter-repetition pauses) and converted to MP3 via ffmpeg. The browser ▶ button plays that file first and uses Web Speech only as a fallback. The CLI defaults to a slower, clearer 130 wpm and routes the English-like respelling through whichever OS TTS backend it detects.

How to contribute

1. Find a word that's missing or wrong.
2. Open a PR adding or fixing one row in pronunciations.tsv.
3. Mark confidence honestly. Leave source_url empty rather than fabricate.
4. ./bin/say-it -o /tmp/test.aiff <word> && afplay /tmp/test.aiff to verify the respelling sounds right.

See CONTRIBUTING.md for details.

What's NOT in scope

• General English vocabulary — there are real dictionaries for that.
• UK / AU / IE / ZA accents. We're General American only, by design.
• Fabricated sources. We'd rather under-claim than make stuff up.

Project

MIT licensed. Maintained on GitHub. The dictionary, the CLI, the MCP server, the website, the OG cards, the audio files — everything is in one repo.

Acknowledgements

Steve Wilhite (GIF), Douglas Crockford (JSON), Linus Torvalds (Linux), Leslie Lamport (LaTeX), Donald Knuth (TeX), Igor Sysoev (nginx), Salvatore Sanfilippo (Redis), Brandon Philips (etcd), Kelsey Hightower (kubectl), Evan You (Vue / Vite), Samuel Colvin (Pydantic), and every other creator whose stated pronunciation we cite. This project is just a thin wrapper around your prior art.