Free and open-source. Record any browser tab — get a clean, speaker-labeled transcript. The only thing you need is a Gemini API key from Google AI Studio (free tier works).
No subscription. No account. No backend. No telemetry. Just a Chrome extension and your own key.
Other meeting transcribers run on remote servers, charge monthly, or require a native desktop app. TabScribe is none of those things.
- 100% free, MIT-licensed. No paid tier, no trial limits, no "premium features". The whole codebase is here.
- Free Gemini tier is usually enough. A free Google AI Studio key handles typical personal use. Heavy users pay Google directly — pennies per hour on
gemini-3-flash-preview. - Local-first. Audio and transcripts live in your browser. The only outbound request is to Gemini, with your own key.
- No install ceremony. It's a single unpacked Chrome extension. No native binary, no build step, no sign-up.
- Readable code. Vanilla JS, no bundler, ~1000 lines total. Fork it, change it, audit it.
That's it. Two things, both free.
- Google Chrome (or any Chromium-based browser that supports Manifest V3 extensions)
- A free Gemini API key from Google AI Studio — sign in with Google, click "Create API key", copy
- Captures audio from the active tab — Meet, Zoom (web), YouTube, podcasts, anything that plays through Chrome
- Mixes in your microphone so your own voice is in the transcript too
- Sends the recording to Gemini for transcription with auto-detected language
- Returns clean markdown with Participants, Transcript, and Summary sections
- Lets you supply a name list and a global dictionary so proper nouns are spelled correctly
This extension isn't on the Chrome Web Store. Install it manually in 30 seconds.
- Clone or download this repo:
git clone https://github.com/<your-user>/tabscribe.git
- Open
chrome://extensionsin Chrome - Toggle Developer mode in the top-right corner
- Click Load unpacked → select the
tabscribe/folder - Pin the TabScribe icon to your toolbar for quick access
Updates: git pull and click the refresh icon on the extension card.
When you first open the popup you'll see a setup checklist. Two steps:
1. Add your Gemini API key
- Get one at aistudio.google.com/app/apikey (free tier works)
- Click the TabScribe icon → gear (⚙) → General tab
- Paste the key → Save
2. Grant microphone access
- Right-click the TabScribe icon → Options (opens a real tab — extension permission prompts only work from visible tabs)
- Click Grant Microphone Access → choose Allow on every visit
- Click Test Microphone to verify it's actually capturing audio
When both checkmarks turn green, the setup banner disappears and the Record button activates.
Recording. Open any tab, click TabScribe → Record tab. Chrome shows a red dot on the tab to indicate capture. You'll still hear the tab audio normally. Click Stop recording when done — transcription starts automatically and lands in the list when ready.
Reading. Click any recording to see the transcript. Copy Markdown drops the full formatted transcript onto your clipboard — paste into Notion, Obsidian, Linear, anywhere.
Continuing. If a meeting got interrupted (closed tab, network drop, you forgot to hit record at the start), open the recording and click Resume recording. New audio appends to the existing one and the transcript is regenerated on the combined file.
Getting names right. Two layers help here:
- Per-recording Participants (in the recording detail view): one name per line. Gemini tries to attribute each spoken line to a real name based on context.
- Global Dictionary (Settings → Dictionary): proper nouns, acronyms, your last name. Sent with every transcription so Gemini spells them right.
| Feature | Where |
|---|---|
| Speaker diarization with real names | Recording detail → Participants |
| Global vocabulary hints | Settings → Dictionary |
| Resume after interruption | Recording detail → Resume recording |
| Re-run transcription with new prompt | Recording detail → Re-transcribe |
| Download raw audio (.webm) | Recording detail → Download Audio |
| Auto-save on tab close | Automatic — no setup needed |
| Bulk cleanup | Settings → Maintenance |
The dropdown is locked to four Gemini 3 models. Pick based on speed/quality tradeoff:
| Model | Best for |
|---|---|
gemini-3-flash-preview |
Default. Fast, cheap, great quality for most meetings. |
gemini-3-1-flash-lite |
Even cheaper. Use for long, low-stakes recordings. |
gemini-3-pro-preview |
Higher accuracy. Slower. Long meetings with many overlapping speakers. |
gemini-3-1-pro-preview |
Latest pro variant. Use when others don't satisfy. |
If your key doesn't have access to a model, it appears greyed out in the dropdown.
- Your audio reaches exactly one external endpoint:
generativelanguage.googleapis.com(Gemini API) - The request uses your own API key — Google bills you, not us
- The extension has no other network access. No telemetry, no analytics, no error reporting
- API key, recordings, and transcripts are stored in your browser's IndexedDB
- Nothing is synced to your Google account or any cloud
- Manifest V3 extension, plain HTML/CSS/JS, no build step
chrome.tabCapture.getMediaStreamIdfor the tab audio streamnavigator.mediaDevices.getUserMediafor the microphone- Web Audio API mixes both into a single
MediaStreamDestination MediaRecorder(WebM/Opus) writes chunks every second- Recording runs in an Offscreen Document so it survives popup closes
- IndexedDB stores audio blobs and metadata
- Gemini REST API: inline base64 for audio ≤ 18 MB, Files API (resumable upload) for larger files
- Works only with tabs in this Chrome window. Native Zoom / Teams / Meet apps don't expose a tab to capture.
- Transcription quality depends on audio quality. Mic gain too high or too low = worse output.
- Concatenated WebM (from Resume) plays fine in modern apps but may glitch in very old players. Gemini handles it without issue.
- Chrome may terminate the Service Worker on long idle. Active recording keeps it alive via the offscreen document.
The codebase is intentionally small and free of frameworks. Pull requests welcome. Ideas: options page UI improvements, transcript editing, export to other formats, content-script integration with Meet/Zoom DOM to auto-fill participant names.
MIT. Use it, fork it, ship your own version — just don't ship it as-is to the Chrome Web Store under the TabScribe name.
