Voice Recorder

Which format you get, and why

Browsers record with the codecs they ship, and they ship different ones. Chrome, Edge and Firefox record Opus audio in a WebM file; Safari records AAC in an MP4 container, saved as .m4a. The recorder asks the browser what it can write and uses the best option, and where a browser can write more than one — recent Safari can do both — it lets you choose.

Opus is the more efficient codec: at 64 kbps speech is clean and natural, which is why calling apps use it. AAC in M4A is the more widely playable file, opening in every Apple app, Windows' media player and most phones. WebM plays in every browser, in VLC and in most modern players, but not in QuickTime or older Windows apps.

Each take downloads as a properly seekable file. MediaRecorder writes WebM as a live stream with no length in its header, which leaves players unable to seek through it, so the recorder writes the duration in before you download. That is a small change to the header; the audio itself is untouched.

WAV for editing, and why it is bigger

Every take can also be downloaded as WAV. The recording is decoded in your browser and written as uncompressed 16-bit PCM, which every audio editor, DAW and operating system opens without a codec — the safe choice for Audacity, GarageBand, a video editor, or a service that asks for WAV.

Uncompressed audio is large. At the usual 48 kHz decode rate, mono WAV is about 5.5 MB a minute and stereo twice that, against under 1 MB a minute for a 128 kbps original. A microphone delivered as two identical channels is written as mono, which halves the file without losing anything.

Converting to WAV does not improve quality: it is the same lossy recording in a lossless container. If quality matters, record at a higher bitrate in the first place, and convert once.

MP3 and other formats

There is no MP3 button here because no browser can record MP3, and bundling an MP3 encoder would add weight to every visit for a format most people do not need. If you need MP3 — for an older car stereo, a podcast host or a form that insists on it — download the take and convert it with the audio converter on this site, which runs ffmpeg in your browser and is just as private.

The same converter turns a take into M4A, OGG or a smaller mono file, and the speech-to-text tool will transcribe one without uploading it.

Getting a good recording

Distance matters more than the microphone. Fifteen to thirty centimetres (six to twelve inches) from your mouth, slightly off to one side, gives a full sound without the pops of breath on p and b. A small room with soft furnishings — curtains, a sofa, a rug — beats a large hard-walled one, because echo is harder to remove than noise.

Watch the meter. For speech, peaks around −12 to −6 dBFS leave headroom without sounding quiet; if the bar turns red, move back or lower the input volume in your system's sound settings. Record a ten-second test first and listen to it on headphones.

Voice processing — noise suppression, echo cancellation and automatic gain — is on by default and helps in a noisy room or with a laptop microphone. Turn it off for music, singing or instruments: it treats sustained notes as noise and makes the volume pump.

Where your recordings are kept

Takes are held in this tab's memory and nowhere else. They are not uploaded and not saved to the site, and they are gone when the tab closes, reloads or navigates away — which is why the page asks before you leave while any exist. Download each one you want to keep. An hour at 128 kbps is under 60 MB, so long recordings are fine; very long sessions are limited by the memory your device can spare.

On a phone, keep the screen on and the browser in front: mobile browsers may pause or stop a recording when the tab goes into the background or the phone locks.