How to extract audio from a video (and which format to pick)

Toolkit Guides · Updated July 2026 · 6 min read

Some of the most useful audio files start life as video: a two-hour lecture recorded on a phone, an interview filmed for the picture but needed for the transcript, a voice memo captured with the camera app because it was the nearest button, a concert clip where the sound is the actual keepsake. Extracting the audio turns a bulky video into a small file you can play anywhere — and it's one of the fastest conversions there is, because the video data never gets touched. The real decision is what to extract to. Here's when extraction makes sense, why it's so quick, and how to choose between MP3, M4A and WAV.

When pulling the audio makes sense

Lectures and talks. A 90-minute recording of a slide deck is mostly a static image. As audio it plays on a commute, at higher speed, without burning storage or battery on video you never look at.
Interviews and meetings. Transcription tools want audio, and a small audio file is far easier to move, archive and share than the video it came from.
Voice memos shot as video. The camera was the nearest button. Keep the sound; let go of two hundred megabytes of pocket lining and ceiling.
Music practice. Pull a song or rehearsal recording into an audio file to loop a section — and if you need the individual parts rather than the mix, an AI stem splitter can take it from there.
Keepsakes. A recital, a wedding toast, a concert: audio for actual listening, while the original video stays archived.

One ground rule before any of it: extract audio only from recordings you made or otherwise have the rights to use — keeping just the sound of someone else's copyrighted video doesn't make it yours.

Why extraction is fast and video conversion is slow

A video file is overwhelmingly video. Ten minutes of 1080p at a typical 5 Mbps is roughly 375 MB, while the AAC audio track inside it, at 128 kbps, is under 10 MB — around 3% of the file. The two kinds of data also cost wildly different amounts of work. Re-encoding video means decoding and re-compressing every frame: ten minutes at 30 frames per second is 18,000 frames of about two million pixels each. That's why converting video takes minutes and warms up your laptop.

Extraction skips all of it. The converter reads only the audio track and discards the video stream unopened — not a single frame is decoded. From there the audio is either copied bit-for-bit into a new container (a “remux,” possible when the format you want matches the codec already inside, as with AAC into M4A) or decoded and re-encoded into your chosen format. Either way, the work is a sliver of what video re-encoding demands, which is why even an hour-long recording gives up its audio quickly.

Try it: drop a video into the video converter and choose MP3, M4A or WAV under “Convert to.” The first use downloads a ~31 MB engine (FFmpeg compiled to WebAssembly) once; the extraction itself runs on your device and your file is never uploaded. MP3 and M4A come out at 192 kbps, WAV uncompressed.

MP3, M4A or WAV: decide by what happens next

There's no universally “best” output — the right one depends on the file's future, not on abstract quality rankings.

	MP3	M4A (AAC)	WAV
What it is	Lossy — the universal veteran	Lossy — MP3's more efficient successor	Uncompressed PCM — the raw samples
Size per minute	~1.4 MB at 192 kbps	~1.4 MB at 192 kbps, slightly better sound	~10 MB at CD quality
Plays on	Practically everything made since the late '90s	Everything modern; occasional friction with old or cheap players	Everything, but the files are heavy
Pick it when	It must play anywhere, no questions asked	You want the best sound per megabyte	You're going to edit the audio next

Two of those numbers deserve a note. WAV's size isn't an estimate: CD-quality stereo is 44,100 samples per second × 16 bits × two channels, which works out to just over 10 MB per minute regardless of what the audio contains. And M4A's edge over MP3 is real but modest at 192 kbps; it grows at lower bitrates, which is part of why phone cameras record AAC in the first place. (MDN's audio codec guide is a good reference on what each codec trades away.)

The WAV recommendation for editing isn't audiophile fussiness — it's workflow. Editors decode compressed audio anyway; starting from WAV means instant scrubbing, clean cuts, and no second generation of loss when you re-save. Extract WAV, edit, and encode to MP3 or M4A once, at the very end.

Bitrate: speech is cheap, music isn't

Bitrate is how many bits each second of audio gets, and speech needs far fewer of them than music. A single voice occupies a narrow slice of the spectrum: 64–96 kbps is intelligible, and 128 kbps is comfortable for lectures, interviews and memos. Music uses the whole canvas — cymbals, bass, stereo width — and gets noticeably rough below 128; 192 kbps is the common sweet spot, with 256–320 as insurance for critical listening.

The video converter extracts at a fixed, sensible 192 kbps. If you want a leaner file for pure speech — say, a semester of lectures — run the result through the audio converter, which encodes MP3 at your choice of 128, 192, 256 or 320 kbps. At 128 kbps, an hour of speech is under 60 MB.

Extraction can't add quality

The audio track inside the video is the ceiling. Extraction hands you exactly what the camera captured — the phone microphone, the wind, the echoey room — and no output setting raises it. Exporting a 128 kbps recording as a 320 kbps MP3 or a WAV doesn't restore detail the original encoder threw away; it makes a bigger, faithful copy of the same sound. (A WAV of a lossy track still has a legitimate use as an editing intermediate — just don't expect it to sound better.)

Re-encoding can, however, subtract. Every lossy-to-lossy pass — AAC to MP3, MP3 to a lower-bitrate MP3 — discards a little more, and the damage compounds. So extract once from the original video, treat that file as your master, and avoid chains of conversions.

A sensible default

If you just want to listen: extract M4A — the audio in most videos is already AAC, so you're staying in its native family — or MP3 if the file must play on absolutely anything. If it's speech and space matters, extract, then re-encode to 128 kbps MP3. If you're going to edit, extract WAV and encode once at the end. And whatever you pick, remember what extraction is: the same recording in a smaller, more convenient body. Nothing is gained, nothing is lost that wasn't already missing — and none of it ever needed to leave your device.