Pull just the audio out of a video

Extracting audio means keeping one of the tracks in the container and discarding the rest. The browser reads the audio track directly and encodes it to MP3 at 192 kbps. Picture, subtitles and chapters do not survive, because an MP3 has nowhere to put them.

The audio is decoded and then re-encoded. What was in the video was already compressed — AAC, in almost every MP4 — so this is a second pass of lossy compression. For speech, lectures and stream archives you will not hear it. If the audio is music you mean to keep, choose WAV in the target list: the same extraction, without the second pass.

Some videos will not give up their audio. That happens when the track uses a codec the browser cannot decode — older AC-3, for instance — and the tool reports it rather than handing you a silent file.

MP3 is written at a fixed 192 kbps. There is no quality slider, because the choice that matters here is the format rather than the number: WAV when the sound matters, MP3 when it does not.

QDoes it work on long videos?

There is no fixed length limit, but the work happens in the browser, so longer means more time and more memory. Past an hour it can fail outright. For recordings of meetings or streams, trim to the part you need with the audio trimmer first, or split the file before bringing it here.

QWhich track does it take when a video has several audio tracks?

Whichever one the browser selects by default — which on a dubbed release may not be the one you wanted. There is no track picker on this page, so if you need a specific one, drop the others in a video editor before converting.

QCan I run the extracted audio through vocal separation?

Yes — take the exported file to the stem splitter. If that is the plan, though, export WAV rather than MP3. The separation model works from fine detail in the input, and detail that MP3 has already discarded is detail the model no longer has.