ihateaudio

Voice Recorder

Record off your mic. It never leaves this tab.

0:00.00
Nothing recorded yet

Your file stays on your device. Always.

How to use it

  1. Click Start recording. The browser asks for microphone permission at that moment and not before. The page does not touch your microphone on load, so nothing is listening until you decide it should be.
  2. Watch the level meter while you talk. Aim for it to sit around two-thirds of the way across on your loudest phrases; a meter that never moves means the wrong input is selected, and one pinned at the far end is clipping.
  3. Pause and resume as often as you need. The timer counts recorded time only, so a pause does not appear in the finished file.
  4. Press Stop. The recording is drawn as a waveform you can play back, scrub and check before you commit to it.
  5. Download it as the original WebM, as MP3, or as WAV. Or hand it straight to the trimmer or the silence remover without saving anything first.

What the browser is actually doing

Pressing record asks the browser for a microphone stream. The browser, not this page, decides whether to grant it, which is why the permission prompt is drawn by the browser chrome, why a page cannot suppress it, and why the recording indicator in the tab strip stays lit for as long as the stream is open. This page closes the stream the moment you press stop, so that indicator goes out immediately.

The stream feeds two things at once. One branch goes to the recorder, which encodes compressed chunks as it goes rather than holding raw audio, so an hour of speech costs a few megabytes rather than several hundred. The other goes to an analyser node that reads the raw waveform sixty times a second purely to move the level meter. It is never connected to your speakers, which is what stops the recording feeding back on itself.

Microphone access is also gated on a secure context. Over plain HTTP the browser will not even offer the prompt, because an unencrypted page could have been modified in transit by anyone between you and the server. HTTPS or localhost, or no microphone.

Getting a level worth keeping

Recording levels are the single most common thing people get wrong, and the mistake is almost always the same one: too quiet, then fixed afterwards by turning it up. Amplifying a quiet recording amplifies the room, the fan and the preamp hiss along with the voice, and none of that can be separated out later.

Aim for the meter to reach around two-thirds on your loudest phrases and to move visibly on quiet ones. Sit about 15 to 20 centimetres from the microphone. Close enough to dominate the room, far enough that plosives do not thump. Speaking across the microphone rather than straight into it helps with those too.

Note that a default browser microphone stream comes with echo cancellation, noise suppression and automatic gain control applied before the audio ever reaches the page. Gain control is the one you will notice: it lifts quiet passages and pulls loud ones down, which flatters a video call and can make a recording breathe unnaturally. Consistent distance and a consistent voice give it less to correct.

What to do with the file afterwards

Keep the original if you only need to archive or send it. It is the actual captured data with nothing re-encoded, and WebM with Opus is markedly better than MP3 at speech bitrates. Choose MP3 when the recipient's software is old or unknown, since it is the one format everything opens. Choose WAV when the recording is going into an editor, because every further edit then starts from lossless audio.

Most spoken recordings benefit from two quick passes before they go anywhere: trimming the dead air at each end where you were reaching for the mouse, and levelling the result so it plays back at a consistent volume. Both are one click away below, and the recording is handed over directly without you needing to save and re-open it.

Questions

Where is the recording stored?
In this tab, in memory, and nowhere else. There is no upload, no account and no server copy, which also means that closing or reloading the tab destroys the recording permanently. Download it before you navigate away.
Why does the browser ask for permission every time?
That is the browser, not the page. Chrome and Edge remember an allow decision per site over HTTPS; Firefox asks each session unless you tick "Remember this decision"; Safari asks per session by default and can be changed under Settings for This Website. Whichever it does, permission is requested only when you press the record button.
Why is the recording a WebM file?
WebM containing Opus is what Chrome, Edge and Firefox produce natively from the recording API. It is efficient, high quality at speech bitrates, and involves no re-encoding. Safari produces MP4 with AAC instead. The MP3 and WAV options here decode that original and re-encode it, which is why they take a moment longer.
It says the microphone is in use by another application.
Some apps take exclusive control of the input device. Zoom, Teams, Discord, OBS and most DAWs do this. Quit the other application entirely rather than just leaving the meeting, then try again. On Windows, check that no background app holds the device under Settings → Privacy → Microphone.
Can it record system audio, or the other side of a call?
No. The microphone API gives a page microphone input and nothing else, by design. A web page that could silently capture your speakers would be a serious security problem. To record something playing on your computer you need a loopback device such as BlackHole on macOS or VB-Cable on Windows, selected as your input.
Is there a length limit?
Not on the recording itself, which streams to compressed chunks and costs only a few megabytes an hour. The limit is playback: to draw the waveform and offer MP3 or WAV, the file has to be decoded into memory, and past roughly forty minutes on a phone or a couple of hours on a desktop that fails. If it does, the original file is still offered for download intact.
Why does my voice sound processed or oddly quiet?
Browsers apply echo cancellation, noise suppression and automatic gain control to a default microphone stream. Gain control in particular rides your level up and down, which is helpful on a call and audible on a recording. Speaking at a consistent distance of about 15 to 20 centimetres gives it less to do.