Extract Audio saves the sound inside an MP4, MKV or MOV as a separate audio file — turning a concert recording into songs, or a video course into audio you can listen to on the way to work. It is the “Extract audio” tab of the audio/video workbench and offers two paths.

Path one, a lossless stream copy, does not re-encode at all: the audio stream is moved as it is into a suitable container — an AAC or ALAC track is saved as .m4a, MP3 as .mp3, Opus as .opus, Vorbis as .ogg, FLAC and AC3 under their own names, PCM as .wav, and rarer codecs land in .mka. Because nothing is decoded it finishes in seconds and sounds exactly like the video. Path two, “Convert to MP3”, re-encodes with libmp3lame at -q:a 2 (about 190 kbps VBR): the widest compatibility, at the cost of one more lossy pass.

1. 01Add a video— drop it into the workbench, select it and switch to “Extract audio”. The panel header shows the detected audio codec, bitrate and sample rate, for example “AAC · 128 kbps · 48.0 kHz”.
2. 02Or take path one, the lossless copy— once detection finishes, the button turns into text naming a concrete extension such as “Extract losslessly as M4A”; click it. Until detection completes the button is greyed out.
3. 03Or take path two, convert to MP3— click “Extract and convert to MP3”, which does not depend on the detection result and works for any video that has an audio track.
4. 04Download— when the job finishes, click Download in the queue at the bottom for <original-name>_audio.m4a or <original-name>_audio.mp3.

Track selection

— only the audio track ffmpeg selects by default is extracted, so a multi-language or multi-track MKV cannot give you a second track

Videos without audio

— when no audio stream is detected both buttons are disabled with the message “No audio track detected”

The lossless copy depends on detection

— if detection fails (a damaged file, say) the container cannot be determined and only the MP3 conversion is available

Fixed MP3 settings

— converting to MP3 has a single setting, -q:a 2; for 128k / 320k go to Audio Format Converter, which also accepts video as input

What it does not do

— it does not cut time ranges (trim in Trim and Cut Video Clips first), change the volume (use Volume Adjust) or remove vocals (use the karaoke mode of Channel Processing)

The video is only read and split inside the browser and is never uploaded to a server; the first use downloads about 31 MB of ffmpeg.wasm engine from a CDN (the site's asset domain or jsDelivr) and that request contains none of your file content.