Pronounce — VS Code
Hover any tech word — hear how engineers actually say it. Backed by the community-curated Pronounce.ai dictionary — 918 sourced entries, every one tagged with confidence (creator-clarified · community-consensus · contested) and a citable URL.
What you get
- Hover over
kubectl, nginx, GIF, YAML, TOML, Ghostty, wagmi, JSON, Affine, Logseq — popup with IPA, English-style respelling, both readings if contested, the source citation, and a 🔊 Play button.
- Speak selection — highlight any word, press
⌘⇧', hear it. Works on words not in the dictionary too (lower-quality TTS fallback).
- Search dictionary — fuzzy-find any of the 918 entries from the command palette (
⌘⇧P → "Pronounce").
- Status bar
🔊 sayit — always-visible reminder; click to speak the current selection.
- Welcome walkthrough — on first install, a 4-step tour shows the hover, the search, and one-click hear-kubectl.
Install
# VS Code (Microsoft Marketplace)
code --install-extension sayit.pronounce
# Cursor / VSCodium / Zed / Gitpod / Theia / code-server (Open VSX — same ID)
code --install-extension sayit.pronounce
Or click through:
Cross-platform as of v0.3. macOS uses the built-in say; Linux uses espeak-ng (sudo apt install espeak-ng / brew install espeak-ng); Windows uses PowerShell's built-in System.Speech.Synthesis. macOS quality is best; the others are functional.
Why
Engineers mispronounce tools all day — kube-cuddle, engine-eks, Gee-son, yaymul, LOG-sek. Webster doesn't cover them; this does. Creator-clarified where possible (Wilhite on GIF, Crockford on JSON, GNU Project's own FAQ, Tom Preston-Werner on TOML, yaml.org's FAQ on YAML, IETF RFC 9000 on QUIC, Mitchell Hashimoto on Ghostty, EWD archive on Dijkstra) and community-consensus otherwise. The extension surfaces all that inline.
Settings
| Setting |
Default |
Purpose |
pronounce.voice |
Samantha |
Any macOS say voice. List with say -v '?'. Recommended: Samantha, Daniel, Karen, Moira, Tessa. |
pronounce.rate |
200 |
Words per minute (80–500). |
pronounce.repetitions |
1 |
Hover plays once; CLI default is 3 (more annoying on hover). |
pronounce.hoverEnabled |
true |
Master switch for hover popups. |
pronounce.hoverOnlyKnownWords |
true |
Set false to also hover unknown words (lower-quality letter-to-sound TTS). |
pronounce.statusBar |
true |
Show 🔊 sayit in the status bar. |
Companion CLI + web + skill + MCP
The same dictionary powers:
brew install say-it # or ./install.sh from the repo
say-it kubernetes # speaks it 3x with cited pronunciation
- Web: pronounce.renlab.ai — full 918-entry browser with Famous Moments + quiz.
- Claude Code skill: triggers on "how do you pronounce X" / "X 怎么读" — the AI plays sound instead of guessing.
- MCP server: drop into Cursor / Claude / any MCP client.
- Raycast / Alfred / Continue: integrations in the main repo.
Contributing
Wrong pronunciation? Missing word? Open a PR against data/pronunciations.tsv in the main repo. PRs flow into the next extension release — npm run build:dict regenerates the bundled JSON.
If this saves you one cringey standup moment — star the repo. It's how more devs find it.
License
MIT.
