Voice recorder. The audio never leaves your device.

At a glance

Why in-browser recording is different

Every other free online voice recorder works the same way: your microphone audio is streamed to their server, stored there, processed there, and handed back to you as a download. That is a reasonable engineering choice and a terrible privacy one. Whatever you just recorded — an interview, a medical note, a client call, a song idea — now exists on someone else's disk, subject to their retention policy and their breach history.

In voice mode this recorder uses the browser's own MediaRecorder API; in studio mode it takes the raw samples through Web Audio. Either way the audio goes from your microphone into memory in this tab, gets encoded by JavaScript running in this tab, and is saved by your browser's normal download mechanism. There is no upload step, because there is no server involved in the recording at all. You can verify this: open your browser's network tab while recording, and you will see nothing being sent.

What to use it for

• Voice memos and quick notes when you are already at a computer and reaching for your phone is slower.
• Interviews and meetings — particularly anything confidential, where uploading the audio to a third party is the thing you are trying to avoid.
• Practice takes and song ideas, then trim them or level them without ever leaving the browser.
• Pronunciation and language practice, where you want to hear yourself back immediately.
• Recording narration for a video, then running it through the podcast prep tool to normalize and encode in one step.

Getting a usable recording

For speech, leave the quality on "voice". Echo cancellation and noise suppression are implemented in the browser's audio stack and are genuinely good. They remove keyboard clatter, room echo and steady background hum before the audio ever reaches the recorder. In this mode the browser compresses the audio while it records (Opus, or AAC in Safari). The page asks for 256 kbps, far above what speech needs; a browser can choose less.

Switch to "studio" for music, singing, or anything you will edit. The processing is tuned for speech and will fight you: noise suppression treats sustained notes as noise, and auto gain pumps the level between phrases. Studio mode turns all three off, asks for both channels of the input (a stereo mic, or both inputs of an audio interface), and keeps the raw samples as 32-bit float at the microphone's own sample rate. Nothing is compressed at any point, so a WAV or FLAC saved from it is exactly what the microphone delivered. The line under the record button says what the browser actually gave you: some built-in microphones are mono whatever you ask for.

Watch the level meter while you talk. You want it moving through the middle of its range, occasionally reaching the upper part, and turning red only rarely. Red means clipping, and clipped audio cannot be repaired afterwards — no amount of processing recovers a squared-off waveform. If it is consistently red, move further from the microphone or turn its input level down in your operating system's sound settings.

If the recording came out too quiet, that is fixable: run it through the amplifier or the normalizer afterwards.

Clean up background noise with AI

Tick "clean up background noise (AI)" before saving and the take goes through RNNoise, the same small neural denoiser as the AI method on the noise reduction page. It keeps the voice and pulls down what is around it: traffic, keyboard clatter, a fan, a dog barking between sentences. It works on the finished recording, in the browser, and the file is saved with "-clean" in its name. Measured on a take recorded through this page with noise bursts and clicks under the speech, the pauses came out 19.9 dB quieter than the same take saved without it, and that plain take already had the browser's own noise suppression. It is for speech: leave it off for music, and other people talking in the room count as voices and stay.

Limits worth knowing

The recording lives in memory until you save it. There is no time limit as such, but a very long session on a machine with little free memory can fail, and closing or reloading the tab loses the recording. For anything longer than about an hour, save in sections.

The permission prompt is the browser's, not ours. If you block it by accident, the site cannot re-ask — you need to re-allow it from the icon in your address bar.

In voice mode Safari records to MP4/AAC rather than WebM/Opus, which the tool handles transparently. Either way the file is decoded and re-encoded to your chosen format before download. Studio mode never uses a codec while recording, in any browser.

Recording quality is bounded by your microphone. A laptop's built-in mic in a hard-surfaced room will sound like a laptop's built-in mic in a hard-surfaced room, and no processing changes that. Getting closer to the microphone is the single biggest improvement available to most people.