Press record. Get an MP3.
Press record · grant permission once · nothing leaves this page
safe · secure · no server · works offline · fast
…or paste a file you copied in Finder, or just start typing at the cursor.
↓ scroll for about & FAQPress record · grant permission once · nothing leaves this page
safe · secure · no server · works offline · fast
…or paste a file you copied in Finder, or just start typing at the cursor.
↓ scroll for about & FAQEvery online voice recorder asks the same thing of you — press the button, talk, and find out afterwards what you got. OmniViewer inverts it. Before you start, each quality preset shows what it costs per hour and what a full two hours would weigh, and once you are running that number counts up beside the clock. A two‑hour interview at the Voice preset is about 55 MB; at Studio it is about 275 MB. That is not a hard question, and no recorder should make you guess.
A browser records into whatever container the platform ships an encoder for:
WebM/Opus nearly everywhere, MP4/AAC on
Safari. Neither is an MP3, which is why so many browser recorders hand you a
.webm and leave the rest to you.
OmniViewer does the last step for you, in the page. When you stop, the capture
is transcoded by a real ffmpeg compiled to WebAssembly —
the same one the site’s
MP3 converter uses, with
libmp3lame doing the encoding. It is about 31 MB and it is
fetched once, the first time you ever stop a recording, and never on
page load. Nothing is uploaded to do it: the encoder runs on your machine, on
your bytes.
Three settings are on by default whenever a web page opens a microphone, and almost nobody is shown them: echo cancellation, noise suppression and auto gain control. They are the right defaults for a video call and the wrong ones for a recording. Echo cancellation subtracts what your speakers are playing and takes much of the room with it. Noise suppression is a denoiser tuned for speech, so it eats rain, applause and guitars. Auto gain control rides your level for you, which is the last thing you want when the level is the recording.
The CAPABILITIES tab shows all three the way OmniViewer’s camera route shows a webcam’s: what was asked for, what the hardware says it can do, and what the track actually got, with every disagreement flagged. A browser is entitled to keep a switch on no matter what you asked, and this is the only place you will find out that it did.
The TRANSFORM tab works on the recording you already have rather than on the live microphone. That choice buys three things: the render is faster than real time, it is exactly repeatable, and it is undoable — the original buffer is never touched, so A/B is a matter of which one you play.
Two of the effects cannot be built out of Web Audio nodes at all. Pitch shifting that holds duration is the obvious one: a browser can change playback rate, which is the chipmunk effect — pitch and length move together. Moving one without the other means stretching time and resampling back, and OmniViewer does it with a WSOLA alignment search so the grains line up instead of cancelling into that phasey, flanged sound a cheap time‑stretch has. Helium goes up seven semitones and every word stays exactly where it was.
The other is the backgrounds. Street, Rain, Café and Stadium are synthesised. Traffic is brown noise with car passes over it, each a band of noise whose filter opens as it approaches and closes as it goes. A café is six voices that are not voices — bands of noise modulated at a syllable rate, which is the cue the ear reads as “people talking”. Rain is bright hiss with individual drops landing on it. None of it is a sample, so there is no megabyte of audio to download and nobody else’s copyright to worry about, and each one is about fifteen lines of arithmetic.
The moment the MP3 exists it is an OmniViewer file like any other, and the hand‑off never touches your disk. Send it to /mp3 EXPORT to convert it to WAV, FLAC, M4A, Opus or Ogg, or to /mp3 STUDIO to cut a shorter clip out of it losslessly — STUDIO copies whole MP3 frames rather than re‑encoding, so a clip you take there has lost nothing on the way. The same viewing engine powers fastjsonviewer.com and hugecsv.com.
No. OmniViewer is a static page with no server-side processing. The microphone is opened by your browser, the audio is captured and encoded to MP3 on your own machine, and nothing is sent anywhere — there is no server to send it to. The microphone is released the moment you stop recording or leave the page.
Two hours in a single take. The limit is enforced rather than merely documented: the recorder stops itself at 2:00:00 and says so, rather than silently truncating. At the Voice preset that is about 55 MB, and at Studio about 275 MB — both shown before you start.
A real MP3. Browsers cannot record MP3 natively — they record WebM/Opus, or MP4/AAC on Safari — so OmniViewer transcodes the capture in the page with a genuine ffmpeg compiled to WebAssembly, using libmp3lame. The encoder is about 31 MB and is downloaded once, the first time you stop a recording, never on page load.
Voice (64 kbps mono) for speech, interviews and memos — it is small enough to email a two-hour recording. Standard (128 kbps mono) is the default and is transparent for a voice. High (192 kbps stereo) if you want the room or an instrument. Studio (320 kbps) is the most MP3 can carry and is worth it only if the source deserves it.
Almost certainly echo cancellation, noise suppression or auto gain control — three pieces of DSP browsers apply to a microphone by default. They are tuned for video calls, so they remove room sound, treat non-speech as noise to be suppressed, and continuously adjust your level. All three can be turned off under Advanced, and the CAPABILITIES tab shows whether the microphone actually honoured the request.
Yes — the TRANSFORM tab has twenty, rendered offline over the finished recording so they are exact and undoable. Helium and Demon shift pitch while keeping the duration, which a browser cannot do on its own. There is synthesised reverb, echo, robot, telephone, megaphone, underwater, vinyl, whisper, alien, reverse, chipmunk and slow-mo, plus street, rain, café and stadium backgrounds that are generated rather than sampled.
Yes. The finished MP3 hands straight off to OmniViewer’s /mp3 EXPORT tab, which converts it to WAV, FLAC, M4A, Opus or Ogg — from the file already in memory, so it is never written to disk or uploaded on the way. /mp3 STUDIO will also cut a shorter clip out of it losslessly, without re-encoding a single frame.
Yes, on any modern mobile browser that supports MediaRecorder — which is all of them. Safari records into MP4/AAC rather than WebM/Opus and OmniViewer transcodes either one to MP3 the same way. The interface is one large button by design, which is as true on a phone as on a laptop.