Voice recorder

How to record audio in the browser

1. 1

   Allow microphone access

   The browser asks once. The level bar starts moving straight away, so you can check you are audible before recording anything.

2. 2

   Click Record, then Stop when done

   Pause and resume as often as you like. Play the take back before you commit to it.

3. 3

   Pick a format and download

   MP3 for something you will send, WAV for something you will edit further.

The level bar is there before you press Record

The single most common way a recording is wasted is not noise or clipping. It is discovering afterwards that the wrong input was selected and the file is forty minutes of silence. So the meter starts as soon as access is granted, before anything is being captured. Say something; if the bar does not move, the microphone in the selector is not the one you are talking into. The bar is on a logarithmic scale, because ordinary speech sits around −25 to −15 dBFS and a linear meter would spend its whole life pinned near zero.

Why the recording is always mono

A single microphone produces one signal. Saving it as stereo does not add anything — it writes the same waveform into two channels and doubles the size of the file for no gain whatsoever. Everything recorded here is saved as one channel, which is also what every transcription tool and every voice-note player expects.

MP3 or WAV

WAV is the raw samples with a 44-byte header: nothing is thrown away, the file is large, and the browser can write it on its own without downloading anything. MP3 is roughly a tenth of the size at 128 kbps, which is more than enough for speech, but it needs the audio engine, about 8 MB, fetched once and cached afterwards. If you are going to edit or transcribe the recording, take the WAV. If you are going to send it to somebody, take the MP3.

HTTPS is not optional

Browsers only expose microphone access on a secure origin, over HTTPS or on localhost during development. On a plain HTTP page the interface that would ask for permission does not exist at all, which is a deliberate decision by browser vendors rather than a limitation here: a page that can silently listen has to be one you can at least verify the identity of.

Where the recording goes

Nowhere. The audio is captured into memory in this tab, encoded in this tab, and handed to you as a download. There is no upload step, no server-side processing and no copy left behind. Closing the tab discards everything that was not saved. The one-hour limit on a single take comes from memory, not policy: uncompressed audio is about 10 MB per minute, and an hour is already close to what a phone browser will hold.

Frequently asked questions