Extract the audio from a video

This tool writes the audio track of your video into its own file and leaves it exactly as it was encoded. Almost every video you will feed it carries AAC audio, so the result is an M4A file containing the same compressed audio, packet for packet. No decoding, no re-encoding, no quality decision to make — and nothing is uploaded.

Extracting is not the same as converting to MP3

Most sites that offer this take the audio out and immediately encode it again as MP3. That is a second lossy compression applied to something already lossy, and the artefacts of the first pass are treated as signal by the second. You cannot get the quality back afterwards, and the file is not smaller in any meaningful way once you pick a bitrate high enough to hide the damage.

Taking the track out untouched avoids the whole question. The audio in your video was encoded once, when it was recorded or exported, and that is the best version that exists. Copying it into its own container gives you exactly that, in a few milliseconds, with no settings to get wrong. A 6-second clip takes about 15 ms here.

The output is an M4A file. That is the same MP4 container you already have, holding only audio — the extension is a convention that tells software what to expect. It plays in Apple Music and QuickTime, in Windows Media Player and the Films & TV app, in VLC, in every browser, on Android, and it imports into every editor and DAW worth using. If a specific piece of software genuinely demands MP3, that is a re-encode, and we have not built it yet rather than pretending a lossy step is free.

What you get, exactly

The page reports the number of audio packets it copied. A packet here is one AAC frame, normally 1024 samples, so a 6-second stereo track at 44.1 kHz is around 260 of them. That number matching the source, together with a file size that matches the audio portion of the original, is how you can tell nothing was re-encoded.

Sample rate, channel count and bitrate are whatever the source had; there is no way for this tool to change them and no reason to. If the video had a 48 kHz mono voice track, that is what comes out. If it had 44.1 kHz stereo music, that is what comes out.

One detail gets reported because it is real and slightly awkward. AAC encoders add a short run of warm-up samples — priming — before the actual content, and a video file marks them as not-for-playback using an edit list. Those samples are preserved here byte for byte, but an audio-only file gives them nowhere to hide, so the extracted track begins about twenty milliseconds earlier than the soundtrack did inside the video. The page tells you the exact figure. It is inaudible on its own and matters only if you are going to line the audio back up against the picture frame-accurately.

When to use this and when not to

This is the right tool when you want the sound that is already there: lifting a lecture recording out of a screen capture, keeping the audio of an interview, saving a voice memo that arrived as a video, or pulling a soundtrack out to edit elsewhere. In all of those cases re-encoding would only make it worse.

It is the wrong tool if you need a format the destination insists on and M4A is not it. It is also the wrong tool if you need to trim, join, normalise or change the volume — those are audio edits, and doing them properly means decoding the track. We have not built those yet. Removing the sound from a video and keeping the picture is the opposite operation and is a separate tool on this site.

And a legal note we would rather state than imply: extracting the audio from a video you do not have the rights to does not become acceptable because the tool is convenient. This exists for your own recordings and material you are allowed to use.

Why doing it in the browser matters

Audio you extract is often the most sensitive part of a recording — a meeting, a consultation, a family message, a phone call caught by a screen capture. Sending it to a server so that a program there can copy a few hundred kilobytes into a different wrapper is a poor trade.

Nothing here leaves the tab. There is no upload, so there is no retention policy to read, no processing licence to accept and no link that can leak. Open the Network tab in your developer tools while you run it and you will see your video in none of the requests, and the Content-Security-Policy header this site sends means the browser would refuse to send it even if a script tried.

The practical ceiling is device memory rather than a policy: the output is built in the tab, so allow for the output plus about a hundred megabytes of working space — comfortably up to 500 MB of input on a desktop, half that on a phone. Because the audio is a small fraction of a video, this tool copes with larger inputs than the others here, and the page warns you when a file is above the guideline rather than letting the tab die.