Wavezard Now Runs in Your Browser, Free

Wavezard used to be a desktop application you downloaded, installed, and paid $19 for. It is now a web page. Open it and start recording. It is free, there is no account, and nothing you record is uploaded anywhere.

What changed

Everything that used to run natively now runs inside a browser tab:

• Whisper transcribes your audio, using WebGPU where your browser offers it
• Silero VAD finds the speech and throws away the silence
• Resemblyzer tells voices apart, so the transcript says who said what
• SQLite, compiled to WebAssembly, stores your meetings on your own device

There is no backend. Not "we don't look at your data" — there is genuinely no server for the audio to reach. The models are downloaded to your machine once and everything after that happens locally.

Why

The honest answer is that almost nobody was installing it.

A desktop app asks a stranger for a lot of trust before they have seen anything work: download an unsigned binary, click past an OS security warning, install it, and then pay. Every one of those steps loses people, and it turns out they lose almost all of them.

A link does not. The browser now does everything the native app needed a native app for — real audio capture, GPU compute, a local database, gigabytes of storage. There is no longer a good reason to make someone install software to transcribe a meeting.

The privacy claim also got easier to believe. Before, "it all runs locally" was something you had to take my word for. Now you can open your browser's network tab and watch: model files come down, and nothing goes up.

New: drop in a recording

You can drag an audio or video file onto the page and get a full diarized transcript back. No microphone, no meeting, no waiting. It reuses the same pipeline as live recording, so speaker labelling works the same way.

If you want to see whether Wavezard is any good, this is the fastest way.

The trade-offs, up front

Two things are genuinely worse than the desktop version, and you should know them before you start:

Capturing the other people in a call needs Chrome or Edge. The browser asks you to share a tab so it can hear its audio, which means a share picker appears when you start a meeting. On macOS it can capture a browser tab's audio but not the whole system. Your own microphone works everywhere, which is all you need for in-person conversations.

First use downloads about 140 MB of models. Once. They are cached afterwards and the app works offline from then on. You can look around while it downloads; only starting a meeting waits.

There is also no local LLM any more. The desktop version shipped a small model for summaries; a 400 MB download in a browser to produce worse summaries than a free chatbot was not a good trade. Instead you can copy a ready-made prompt into whatever you already use, or plug in your own API key.

What happens to the desktop app

It is discontinued. Nobody bought a licence, so nobody is owed anything, and the browser version does more than it did.

Try it

Open Wavezard. No download, no sign-up, no payment.