How to get the audio out of a video

The sound in a video is already a finished audio file, packed in next to the picture. Getting it out is a matter of copying that stream into a file of its own. You don't need to convert anything unless you want a different format, and copying is instant and loses nothing.

ffprobe -v error -select_streams a -show_entries stream=codec_name,sample_rate,channels -of default=noprint_wrappers=1 video.mp4

codec_name=aac
sample_rate=44100
channels=2

ffmpeg -i video.mp4 -vn -c:a copy audio.m4a

Pick the file type from the codec

Copying changes only the wrapper, so the wrapper has to accept the audio that is already there. This is what FFmpeg 8.1.2 did with -vn -c:a copy on our test files:

The .aac row is a quiet one. An MP4 carries a note telling the player to skip the encoder's 1,024 samples of start-up padding, and a bare .aac file has nowhere to keep it. The copy is still lossless, but it starts 23 ms late, which matters if you line it up with anything. Use .m4a.

Already on a Mac: QuickTime Player

Open the video in QuickTime Player and choose File › Export As › Audio Only. Apple's export guide describes the result as an MPEG-4 audio file with an AAC track, so you get an .m4a.

We didn't click through the menu for this guide. We did run avconvert, the command-line exporter that ships with macOS, with its Apple M4A preset on macOS 26.5.1. For the AAC clip it kept the audio as it was: the .m4a decoded to the same samples as the ffmpeg copy. It couldn't open the WebM at all, so expect the same of QuickTime for a WebM with Opus audio: use ffmpeg or the browser tool.

When you want MP3 or WAV

Sometimes the copy isn't what you need: an old car stereo wants MP3, or an editor or transcription app wants WAV. Then re-encode, from the video itself rather than from a copy you already compressed:

ffmpeg -i video.mp4 -vn -c:a libmp3lame -q:a 2 audio.mp3   # MP3, variable bitrate
ffmpeg -i video.mp4 -vn -c:a pcm_s16le audio.wav           # WAV, 16-bit
ffmpeg -i video.mp4 -vn -c:a flac audio.flac               # FLAC, lossless

The same 8-second clip, whose audio is 131 kbps AAC, came out like this:

A higher bitrate can't restore what the first compression threw away, so a 192 kbps MP3 of 131 kbps audio is bigger without being better. WAV and FLAC keep everything that is left, at eight to ten times the size. Which format to pick for which job is in How to convert audio and video formats.

More than one audio track

A film or a recording from a game capture can carry several tracks: languages, a commentary, the microphone on its own. ffprobe lists them in order. Pick one with -map 0:a:N, counting from 0:

ffmpeg -i film.mkv -map 0:a:1 -c copy commentary.opus

Without -map, ffmpeg takes the one it thinks is best (usually the one with the most channels) and ignores the rest. The browser tool does the same, and has no way to pick another.

If there is no audio at all

If ffprobe prints nothing, the video has no sound to extract. Screen recordings are the usual case: the recorder had system audio and the microphone switched off. ffmpeg says "Output file does not contain any stream". The browser tool currently reports only "Something went wrong while extracting the audio" for such a file, and Try again won't help. Record it again with sound on.

In the browser

Every format here is a re-encode, including M4A and AAC from a video that already has AAC audio. For a lecture or a voice memo the second compression rarely matters. For music you care about, choose FLAC, or use the -c:a copy command above.