Extract audio from a video on a Mac without re-encoding it

You have a video and you want the sound out of it. Three ways to extract audio from video files on a Mac, two of them already installed, and one that does it without touching the audio at all.

Everything below was run on this machine and the numbers are what it printed.

The fastest way, if you have ffmpeg

ffmpeg -i video.mp4 -vn -acodec copy audio.m4a

-vn drops the video, -acodec copy takes the audio track out without decoding it. This is the only way to get audio from video with nothing lost along the way. Nothing is re-encoded, so nothing is lost, and it finishes almost instantly — 0.018 seconds on my test file.

Is it really untouched? I checked, rather than assuming: hashing the audio stream in the source video and in the extracted file gives the same MD5, 1ec36ebf851caaec8dd2ac1d5ead951b. Byte for byte, the same track.

That matters because audio in a video is already compressed, usually AAC. Re-encoding it means lossy compression applied twice, on top of a codec that was tuned for the first pass. Copying skips the question.

The container has to be able to hold the codec — AAC into .m4a works, AAC into .wav does not, because WAV holds uncompressed audio. If ffmpeg complains about the codec, that is what it means.

The way that needs nothing installed

macOS ships afconvert in /usr/bin, and it reads video files directly:

afconvert -f WAVE -d LEI16 video.mp4 audio.wav

This one re-encodes — it decodes the AAC and writes uncompressed 16-bit WAV — but it needs no Homebrew, no download and no permission. It took 0.017 seconds on the same file.

For M4A instead of WAV:

afconvert -f m4af -d aac video.mp4 audio.m4a

afinfo audio.wav will then tell you exactly what you got — sample rate, channels, duration — which is worth a look the first time.

The way with no terminal at all

QuickTime Player: open the video, then File, Export As, Audio Only.

Apple’s guide describes what you get: “Audio Only: An Apple MPEG 4 audio file with an AAC audio track.”

Two clicks, no commands, and fine when you need to extract sound from video once. For twenty files it is twenty rounds of the same menu. That is the point at which the command-line versions start to look attractive.

One boundary worth naming, since QuickTime appears above: it exports the audio of a file happily, and cannot capture the audio your Mac is playing at all — why that is, with the device list.

What the sizes look like

The same audio, three ways, from one small test video:

File	Size
Original video (H.264 + AAC)	74.8 KB
Audio copied out as .m4a	66.4 KB
Same audio as WAV	230 KB

The WAV is three times larger than the entire video it came from, video track included. That is not a mistake — uncompressed audio is simply big, and an hour of it runs to hundreds of megabytes.

So the default should be copying the original track. Convert to WAV only when the next tool in your chain demands it.

Something demands WAV, usually

The reason so many guides tell you to extract to WAV is that command-line speech recognition wants it. Whisper run by hand expects 16 kHz mono WAV, so the tutorial says ffmpeg -i video.mp4 -ar 16000 -ac 1 audio.wav and everyone copies that line without asking whether their case needs it.

Editors mostly do not: Final Cut, Premiere and DaVinci all take an M4A. Neither do most players. It is specifically older audio pipelines that want uncompressed input.

Down-sampling to 16 kHz is worth understanding rather than copying: speech recognition models are trained at that rate, so feeding them 48 kHz gains nothing and the resampling happens anyway. It is not damage, it is matching the model.

When you can skip the whole thing

If the reason you are extracting audio is to get text out of a video, the step is probably unnecessary.

Transcription apps that open video files read the audio track from inside the container themselves. Kekoso, which I build, takes MP4 and MOV directly, and so do most of its competitors. The extraction step in that workflow exists only because command-line Whisper needs it.

The full route from a video file to a transcript — and which formats go in — is a different page. This one is about the audio file, for when the audio file is what you actually want: extract music from video for a soundtrack, pull a podcast edit out of a recording, rescue a voice memo hiding inside a screen capture.

Commands, timings and sizes above were measured on macOS 26.5.2 on 5 September 2026 with ffmpeg 8.1.2. Apple’s description of the QuickTime export is quoted from its own guide.

Questions people ask

How do I extract audio from a video on a Mac for free?

Three ways, two of them already installed. QuickTime Player: File, Export As, Audio Only — Apple's guide describes the result as 'An Apple MPEG 4 audio file with an AAC audio track.' The built-in afconvert command reads video files directly: afconvert -f WAVE -d LEI16 video.mp4 out.wav. And ffmpeg, if you have it, can copy the track without re-encoding at all.

Can I extract audio without losing quality?

Yes, with ffmpeg -vn -acodec copy. That takes the existing audio track out of the container without decoding or re-encoding it — I checked by comparing MD5 hashes of the track before and after, and they match exactly. Any method that re-encodes puts a second round of lossy compression on audio that was already compressed once.

Does macOS have a built-in audio extractor?

Two. QuickTime Player has Export As, Audio Only in the menu. And afconvert, a command-line tool in /usr/bin, reads MP4 and MOV directly and writes WAV, AIFF, M4A and more. Neither needs installing, and on my machine afconvert pulled the track out of a test video in 0.017 seconds.

Why is my extracted WAV so much bigger than the video?

Because WAV is uncompressed. In my test the video was 74.8 KB, the copied AAC track 66.4 KB, and the same audio as WAV 230 KB — over three times larger than the whole video it came from. If you do not specifically need uncompressed audio, copying the original track keeps the file small.

What format should I extract to?

Copy the original codec if you can — usually AAC in an .m4a container — because it is both the smallest and the only lossless option. Convert to WAV only when the next tool requires it, which older speech recognition pipelines often do. Convert to MP3 only if something specifically needs MP3.

Do I need to extract audio before transcribing a video?

Not with a transcription app that opens video files, which most now do — the audio track inside the container is what gets read, and the extraction step buys nothing. It is required when running Whisper from the command line, which wants a 16 kHz WAV, and that is the main reason the step still exists in most tutorials.