Convert video to MP3 or extract the audio

Pulling audio out of a video usually costs nothing

A video file is a container holding separate tracks: one for the picture, one for the sound, sometimes subtitles alongside them. The soundtrack is already a finished, compressed audio file in there — it just happens to be stored next to a video. Extracting it, in the strict sense, means copying those bytes into a file of their own. No decoding, no re-compression, no decision about quality, because none of that is needed to move data from one container into another.

That's what the M4A option does when you leave the quality on “Keep as it is”. Almost every MP4 and MOV carries AAC audio, and .m4a is simply AAC in its own container, so the track lands in the new file byte-identical to the one in the video. A two-hour recording comes out in about as long as it takes to read the file from disk, and what you get is not close to the original — it is the original.

Conversion is a different thing, and this page keeps them visibly separate rather than calling both “extract”. Asking for MP3, or asking for a specific bitrate, means every sample has to be decoded and compressed again. That takes real time and costs a generation of quality. The panel above the button says which of the two you're about to get, on this particular file, before you press anything.

Copying is only possible when what you asked for matches what's already inside: AAC out of AAC, MP3 out of MP3. Anything else is a genuine conversion. That's why the “Keep as it is” option disappears on files it can't apply to — it isn't a slightly worse choice there, it's a thing that doesn't exist.

MP3 or M4A — which one you actually want

MP3 is thirty years old and that is its entire advantage: there is essentially no device that won't play one. Car stereos from 2006, gym equipment, cheap voice recorders, DJ controllers, some editing software with narrow import filters. If you're handing the file to one of those, MP3 is the correct answer and the fact that it re-encodes is just the price.

For everything else, M4A is the better file. It's AAC, which is what your phone, your camera and virtually every video on the internet already recorded — so extracting to M4A is a copy rather than a conversion. It also sounds better than MP3 at the same bitrate, by a margin wide enough that 128 kbps AAC is generally reckoned to match around 192 kbps MP3. Everything made in the last fifteen years plays it, including every browser, every phone, and iTunes-era hardware.

The case worth avoiding is converting an ordinary MP4's soundtrack to MP3 out of habit. That audio is AAC and already lossy; encoding it again as MP3 doesn't recover anything, it throws away a second round of detail and frequently produces a larger file than the one you started with. If nothing on the receiving end requires MP3, extracting the M4A is faster, smaller and lossless.

The MP3 encoder here is LAME, compiled to run in your browser. It's downloaded only if you actually ask for an MP3 conversion — picking M4A, copying, or removing the sound never fetches it.

Bitrate, and how small audio can honestly get

Bitrate is the one setting here, and it's unusually straightforward: it's a promise about size. 192 kbps means 192 kilobits for every second of audio, so an hour of it comes to roughly 86 MB no matter what the recording contains. That arithmetic is why audio compression is predictable in a way video compression never is — pick a bitrate and you have effectively picked the file size.

What the number buys you is detail, and the useful range is narrower than the options suggest. For speech — interviews, lectures, podcasts, voice memos — 96 kbps is plenty and 128 is generous; going higher spends bytes on hiss. For music, 192 is the point where most people stop being able to pick the encoded version out of a comparison, and 256 or 320 is for archiving or for material you expect to be re-encoded again later. Below 96, music starts to sound obviously thin, especially cymbals and applause.

The lists differ between the two formats on purpose. MP3 offers 96 through 320; AAC starts lower, at 64, because it stays usable further down. If your goal is simply the smallest listenable file — a two-hour meeting recording you want to keep, say — 64 kbps AAC does that job at about 29 MB an hour, and MP3 has no equivalent.

Every bitrate other than “Keep as it is” re-encodes, including the highest one. Picking 320 kbps on a 128 kbps source doesn't restore anything — it stores the same audio less efficiently. When you want the file left alone, the option that does that is “Keep as it is”.

Removing the sound from a video

The third option inverts the other two: keep the picture, throw the audio away. It's a real removal rather than a volume change — the track isn't written into the output at all, so there's nothing to turn back up and the file is slightly smaller for its absence. The picture is untouched, and when the source is already an MP4 or MOV nothing is decoded either, so this finishes in seconds regardless of length.

The most common reason to do it has nothing to do with editing. Screen recordings pick up whatever was being said in the room; phone videos of a school play capture the conversation of whoever was standing behind you; a clip you want to send captured background music you'd rather not pass on. Stripping the audio before the file goes anywhere is a privacy operation, and it's a much better one when the file never leaves the machine to have it done.

One case does cost time. If your source is WebM or another format built around VP9 or AV1, producing an MP4 means re-encoding the picture into H.264, because those codecs can't simply be repackaged into an MP4 container. The panel says so before you start, and it's the only situation where removing the sound is slower than a few seconds.

Surround, Dolby, and unusual sample rates

Copying a track out preserves everything about it, including 5.1 or 7.1 channel layouts. Re-encoding can't: LAME writes at most two channels, and the AAC encoders browsers provide reject multichannel input outright rather than downmixing it themselves. So any conversion here folds surround into stereo and the panel says so before you start. If you need the channels kept, copying is the option that does it.

Dolby Digital is the harder case. AC-3 and E-AC-3 tracks can be copied — they land in the output whole, and desktop players handle them fine — but they cannot be re-encoded, because no browser ships a Dolby decoder to read them with in the first place. Rather than exporting a silent file and letting you discover it later, choosing a bitrate on a Dolby track is blocked outright, with “Keep as it is” offered as the way through. One consequence worth knowing: a copied Dolby track may be silent if you preview the downloaded file in a browser tab, for exactly the same reason. It isn't damaged; the browser just can't decode it.

Sample rate is the last of these. MP3 accepts only nine of them, an artefact of the MPEG specifications it was standardised in, and high-resolution audio recorded at 96 kHz isn't among them. Those files are resampled to 48 kHz on the way out — the same frequency family, half the rate, and comfortably beyond the range of human hearing either way. AAC has no such restriction, so extracting to M4A keeps the original rate.

What this tool doesn't do

This is a single-pass extraction tool rather than an audio editor, and the boundaries are worth stating outright instead of leaving you to find them.

1. It writes M4A, MP3 and MP4 only. There's no WAV, FLAC, OGG or WMA output.
2. One file, one pass. No batch queue, no joining two recordings together, no extracting several tracks from a file that has more than one.
3. No trimming, no fades, no volume normalisation and no noise reduction. To cut a section out first, use the adjust tool, then extract from the result.
4. Adding or editing track titles, artists and cover art isn't part of this.
5. There's no preview player. The plan panel describes what you'll get; listening happens after you download.
6. There's no size limit, but there is a disk check. The output is written to local storage as it's produced, so memory doesn't grow with the file — but the space has to exist, and it's verified before anything starts.

Your original is never modified or moved. What you download is a new file, so deleting the source stays a decision you make yourself, afterwards.

FAQ