Voice to Text

Free voice dictation in any text field, on any website. Click into a text box, click the microphone that appears, and speak — your words are transcribed live into a small popup and typed straight into the field. 20+ languages, no account, no API key.

Built for high-quality Persian (فارسی) dictation, with first-class support for English and 18 other languages.

Status: v0.1.0 — works end to end. Not yet on the Chrome Web Store; install from source (below).

---

Features

• Click-to-dictate anywhere — a mic icon appears at the corner of any focused text field (input, textarea, or contenteditable). No per-site setup.
• Live transcription — see interim results as you speak in an attached popup, with a real-time audio-level meter so you know the mic is hearing you.
• Types into the field and shows the popup — or switch to popup-only mode and copy manually. Your choice.
• 20+ languages — English, Persian, Arabic, Turkish, German, French, Spanish, Italian, Russian, Portuguese, Hindi, Japanese, Korean, Chinese, Dutch, Polish, Indonesian, Swedish, Ukrainian, and more.
• Smart, per-language output — numbers and punctuation are written the way the language expects, automatically: Persian → ۱۲۳ and نیم‌فاصله, Arabic → ٠١٢٣, everything else → Latin. No toggles to fiddle with.
• Spoken punctuation commands — say "period", "comma", "new line" (or "نقطه"، "ویرگول"، "خط جدید"، …) in any supported language to insert punctuation.
• Fully translated UI — pick the extension's own language independently of the language you speak. Right-to-left layouts for Persian and Arabic.
• Keyboard shortcut — Ctrl+Shift+Y (⌘⇧Y on macOS) starts/stops dictation on the focused field.
• Per-site on/off — silence the mic icon on any site from the toolbar popup.
• Light / dark / auto theme.

How it works

1. On first install, an onboarding page opens and asks for one-time microphone access.
2. Click into any text field on any page — a small mic icon appears just outside its top-right corner.
3. Click the icon (or press the shortcut). It turns into a stop button and a popup opens, transcribing live.
4. Click stop to end. Copy the text from the popup, or let it type straight into the field.

Install from source

Requires Node.js 18+.

npm install
npm run build

Then load it into Chrome:

1. Open chrome://extensions.
2. Enable Developer mode (top-right).
3. Click Load unpacked and select the .output/chrome-mv3 folder.

For live-reload development instead of a one-off build:

npm run dev

Settings

Open the settings from the toolbar popup (⚙️) or the onboarding page. Everything auto-saves:

Setting	What it does
Speech language	The language you speak. Drives recognition, digits, and punctuation.
Extension language	The language of the UI itself (independent of speech).
Theme	Auto (match system), light, or dark.
Text insertion	Type into the field + popup, or popup-only.
Spoken punctuation	Turn "period"/"نقطه"/… into real punctuation.
Per-site	Disable the mic icon on the current site (toolbar popup).

Privacy

• The extension has no servers, no account, and no analytics. Your settings live only in your browser's local storage.
• Recognition is powered by the browser's built-in Web Speech API. Be aware that in Chrome this engine currently streams audio to Google's speech service for processing — that is Chrome's implementation, not something this extension adds. No audio is sent anywhere by the extension itself, and nothing is stored.
• The microphone is only active while you are dictating.

Permissions

Permission	Why it's needed
offscreen	Runs the speech-recognition engine in an offscreen document (the only place an extension can hold a live microphone stream in Manifest V3).
storage	Saves your settings locally.
<all_urls> (host)	So the background service worker can deliver transcription results back into the content script on whatever tab you're typing in. (runtime.sendMessage can't reach content scripts; results must go through tabs.sendMessage, which needs host access.)

Microphone access itself is not a manifest permission — it's governed by the standard browser getUserMedia origin-permission model, granted once during onboarding.

Known limitations

• Google Docs — the Docs editor draws text on a <canvas> rather than into a real editable element, so the mic icon can't attach to it. Dictate into the popup and copy, or use it in any normal field elsewhere.
• chrome:// pages, the New Tab page, and the Chrome Web Store — browsers forbid extensions from running on these pages, so the mic icon won't appear there (including Chrome's default new-tab search box).
• Closed shadow DOM — fields inside a site's closed shadow root are invisible to any extension and can't get an icon. Open shadow DOM (the common case — YouTube, Google Translate, most web components) works fine.
• Recognition accuracy depends on the browser's built-in speech engine, your microphone, and background noise — it varies by language and accent.

Tech stack

• WXT + TypeScript, Manifest V3
• Chrome Offscreen Documents for the mic + webkitSpeechRecognition engine
• Shadow DOM for the on-page UI (isolated from host-page styles)
• @persian-tools/persian-tools for Persian normalization
• Vazirmatn font (embedded) for Persian/Arabic UI

Development

npm run dev        # dev build + live reload (Chrome)
npm run build      # production build → .output/chrome-mv3
npm run compile    # type-check only (tsc --noEmit)
npm run zip        # package for distribution

A note on npm audit

npm audit reports advisories in the development toolchain (wxt → web-ext-run → fx-runner/node-notifier/tmp/uuid). These are build/dev-time only. npm audit --omit=dev reports 0 vulnerabilities — nothing in the shipped extension is affected. Do not run npm audit fix --force; it would downgrade/break the WXT build.

Contributing

Translations especially are welcome. UI strings for the widely-spoken languages are hand-written; the less common ones are best-effort. They live in utils/i18n-messages.ts — each language must provide every key (TypeScript enforces this), so corrections and improvements are easy to spot and PR.

License

MIT © 2026 Sam Nodehi.

Bundled third-party components (Vazirmatn font, persian-tools) retain their own licenses — see THIRD_PARTY_NOTICES.md.