Free meeting transcription for your meeting tab and microphone.
Keep, search, and revisit your words. English speech, processed on your device.
I’ll update the onboarding checklist before Friday. Can we send it with the launch email?
Yes. Let’s include the setup guide, too. That was the question we heard most often in testing.
Google Meet/
Zoom in your browser/
Microsoft Teams/Voice notesShare a tab with audio in a supported desktop browser. Check your setup →
The name, the deadline, the “let’s do it this way.” Give those details somewhere to live.
Share a supported meeting tab, use your microphone, or combine both. No bot joins the conversation.
Your device does the listening. Meeting audio never needs to visit a transcription server.
Search titles and transcript content across your personal library. Less digging, more finding.
Copy a transcript or export it as text, Markdown, or structured JSON. Your words aren’t locked in.
Sign in with Google for your own private transcript library.
Choose a browser tab and turn on audio sharing. Let local AI get to work.
Stop, review, and save. Come back to your words whenever you need them.
KeepMyTalk is a free browser-based AI meeting transcriber for English speech. Capture a meeting tab, your microphone, or both. Save, search, edit, and export transcripts.
Your device does the transcription with open-source AI, so no paid speech API is required.
Check browser and device requirementsLocal processing uses your device. Hosting and storage quotas still apply.
Meeting audio + microphone
Local speech recognition
Search, edit, export, delete
Your audio is processed on your device with open-source AI. Only transcript text and meeting metadata are saved to your account. No audio archive. No hidden meeting bot.
Read how we handle your dataYes. KeepMyTalk is free to use, with no subscription or credit card required. Meeting transcription, your transcript library, search, editing, and exports are included. Your device runs the speech model. Hosting and database quotas still apply; free usage does not mean unlimited capacity or guaranteed availability.
No browser extension is needed. After sign-in, about 68 MB of default Fast speech files download in the background to this browser. Local AI initializes on New transcription. The optional Detailed model downloads when selected (roughly 105 MB including runtime and filter). A recent desktop browser and a capable laptop are recommended.
Share the browser tab containing your meeting and enable audio in the browser picker. Browser-based meetings work when an audio track is provided. Capturing a native desktop meeting depends on your browser and operating system.
Tab audio is the recommended path in desktop Chromium browsers. Whole-screen and window audio vary by operating system. Microphone capture is a separate option. See Help for testing status.
Yes. Choose microphone only, tab audio, or both. Use headphones when combining sources to reduce echo and duplicate speech.
You can sign in, read, search, edit and export transcripts on a phone. If your browser supports secure microphone capture and AudioWorklet, microphone mode is available; meeting-tab and other app audio are not a reliable mobile workflow. Keep the recording page open and the screen awake. Local AI speed and memory requirements vary, and physical iOS/Android recording remains unverified. Try a short transcription first, or use desktop Chrome or Edge for meeting tabs.
The model may be cached after downloading, but signing in, loading the application, and saving to your account require a connection. Offline application startup is not supported.
Text, timestamps, and meeting metadata are stored in the deployment’s PostgreSQL database and scoped to your account. Audio is processed in your browser. Opt-in pending audio is temporarily buffered on your device and deleted after recognition; it is never uploaded.
Export each transcript as TXT, Markdown, or JSON. Delete individual transcripts in the library, or delete all transcripts and your account in Settings.
Draft recognition starts with about 1.5 seconds of captured audio, plus device processing time. Draft words may change as more context arrives; only finalized text is saved. Busy devices update drafts less often to prioritize final text. If recognition falls behind, listening continues while pending audio is buffered on your device. Storage exhaustion or browser/device interruptions can still stop capture with an explanation. Accuracy is imperfect; review important details.
Completed text is saved incrementally. A temporary, account-specific recovery copy is kept in this browser until saving succeeds. Committed pending audio and checkpointed text can be recovered from New transcription in the same browser and account. The last partial window may be lost. Sign-out, private-mode closure or clearing site data removes local recovery.
Give your next conversation a place to land. Save the transcript, revisit the details, and keep what’s useful.
Start for free Free to use. No subscription. No credit card.