AI-PoweredMeeting NotesThat Never Leave Your Browser

Wavezard is an AI notetaker that records, transcribes, identifies speakers, and summarizes your meetings. It runs entirely inside your browser tab. No bots join your calls and no audio is uploaded anywhere.

The models run on your own machine, in the tab. Whisper handles speech recognition, Silero detects speech, and Resemblyzer tells voices apart. They are downloaded once on first use, cached by the browser, and everything after that is local. There is no Wavezard backend for your audio to reach.

No. Open the app and start. There is no download, no installer, no sign-up and no account. Closing the tab and clearing site data is the uninstall.

Wavezard asks your browser to share your audio, so you will see the share picker when you start a meeting. On Windows and Linux, pick Entire screen and tick Share system audio at the bottom-left - that hears everything, including native apps like Zoom and Teams, not just a browser tab. On macOS the picker cannot carry whole-screen audio at all, so pick the tab your call is in and tick Share tab audio instead. This is also why Wavezard requires a Chromium browser: Firefox and Safari return video with no sound.

About 130 MB in total for the default English model, plus 8 MB for voice activity and speaker recognition. It happens once and is then cached. You can explore the app while it downloads; only starting a meeting waits for it. A smaller and a larger model are both available in Settings.

Yes, with the multilingual model, which you can pick in Settings. Whisper claims 99 languages. The default English-only model is smaller and faster; the multilingual one is a bigger download and needs more processing power.

A Chromium browser on a desktop - Chrome, Edge, Brave, Opera, Vivaldi or Arc - and roughly 200 MB of free storage for the models and your meetings. Wavezard will not open in Firefox or Safari: hearing the other people in a call depends on sharing a tab's audio, and those browsers cannot do it, so a meeting would capture your voice and nobody else's. WebGPU, which Chromium has on most modern hardware, makes transcription several times faster.

In your browser's local storage, on this device only. You can export any meeting as CSV or JSON, and export the entire database as a single file to back up or move to another machine. Nothing is synced and there is no copy anywhere else, which also means clearing your browser data deletes it.

Nothing. Wavezard is free. There is no licence, no trial, no subscription and no paid tier. Summaries can optionally use your own API key with a provider you choose, which is billed by them, not by us; there is also a copy-the-prompt option that needs no key at all.

Silero VAD for voice activity detection, Whisper for speech recognition via transformers.js, Resemblyzer for speaker identification via ONNX Runtime Web, and SQLite compiled to WebAssembly for storage.