How to extract audio from a video (MP3, WAV, AAC)

How do I extract audio from a video for free in my browser?

Use the Koldflux converter. It runs ffmpeg inside your browser tab, so the video stays on your device. The tool pages state a 100 MB limit per file and a one-time download of about 30 MB for the converter engine on the first conversion. We ran it end to end on October 21, 2026 (details in the test-file section below): a 10 second MP4 became a playable 128 kbps stereo MP3 in about 4 seconds. The output settings listed below are what the tool pages state, and our own ffmpeg runs further down show what those settings produce.

1. Open the converter that matches your file, for example MP4 to MP3, MOV to MP3 or MP4 to WAV. The full list is on the Video to Audio page.
2. Choose the file. Wait for the engine to load the first time.
3. Start the conversion. The picture is dropped and only the sound is kept.
4. Download the audio file. MP3 output is stereo at about 128 kbps, WAV is 16-bit PCM, AAC is about 128 kbps, and M4A is AAC in an audio-only container (as stated on each tool page).

The tool is a fit for short and medium files. For a file over 100 MB, such as a long recording, use ffmpeg below, which has no such limit.

Free tools:Video to audio converters (all formats)MP4 to MP3MP4 to WAVMP4 to AAC

How do I extract audio with ffmpeg?

ffmpeg is free and open source and works on Windows, Mac and Linux. The flag -vn means "no video". We ran each command below on October 21, 2026 with ffmpeg 8.1.2.

Video.mp4 with AAC sound.

Copying (-c:a copy) is the only route that loses no quality, because nothing is re-encoded. It only works when the audio codec fits the new container: AAC into .m4a is fine; to get MP3 or WAV you must re-encode.

Which format should I pick: MP3, WAV or AAC?

Pick by what you do next. We extracted the audio of one 30 second test video (generated, AAC 192 kbps, 44.1 kHz stereo) six ways and measured the file sizes.

Measured on October 21, 2026 with ffmpeg 8.1.2, 30 second clip, size per minute is bytes times 2.

Method: one generated test video (a test pattern with mixed tones and light noise), audio AAC 192 kbps 44.1 kHz stereo, extracted with the commands above. WAV and fixed-bitrate sizes depend only on length; variable-bitrate (VBR) sizes depend on the sound, and our tones are simpler than music, so real music usually gives a larger VBR file. Treat the VBR row as a floor, not a promise.

Short guide: WAV when you will edit or mix the sound further (no lossy step added, but big). MP3 when you need the widest compatibility and small size. AAC or M4A when the video already has AAC sound: copy it and lose nothing. A lossy-to-lossy conversion (AAC video to MP3) always adds a second generation of loss, so do not go lower than the source bitrate. Converting to WAV does not bring back anything the AAC or MP3 already threw away: it only stores the already-decoded sound without a new lossy step, which is why a WAV from a 128 kbps source is large but sounds the same as the source. A high bitrate alone does not guarantee that you cannot hear a difference either; what you hear depends on the sound, the encoder and your speakers, and we measured sizes and levels, not listening quality. FFmpeg's MP3 guide says the quality settings -q:a 0 to -q:a 3 normally give transparent results. For a longer comparison see the MP3 vs WAV vs FLAC guide.

How do I extract audio on a Mac with QuickTime Player?

Apple's QuickTime Player guide ("Export movies") says to choose File, then Export As, then pick Audio Only. The result, according to Apple, is an MPEG-4 audio file with an AAC audio track (an .m4a file). Apple's page shows no date other than 2026 copyright; we did not run macOS steps ourselves for this page, apart from the ffmpeg runs above.

How do I separate audio from a video in Clipchamp on Windows?

Microsoft's Clipchamp page on muting, separating and deleting audio says to select the video clip on the timeline, open the Audio tab in the property panel, and use the detach control. The audio then appears as its own clip, and Microsoft states that the detached audio file also downloads to your computer automatically. The page does not name the file format, so check the extension after download. That page has no date besides the 2026 copyright, and we did not run it ourselves.

What about VLC?

VLC can convert a video to audio (Media, Convert / Save), but we did not find current official VideoLAN steps for audio-only output: the VideoLAN wiki page we could reach was last edited on March 2, 2019 and says it is outdated. Until we can verify the steps on the current version, use the browser tool, ffmpeg or the app routes above.

Can I check the converter with a test file?

Yes. We made a 10 second sample video with ffmpeg (a test pattern with a 440 Hz tone on the left and a 660 Hz tone on the right, AAC 128 kbps, 44.1 kHz stereo, 320 by 180). Download it with the button under the player, run it through the converter, then compare your output with the numbers below using ffprobe -show_entries format=duration,size on your file. It is our own file, free to use for this test.

What happened when we did it on October 21, 2026: we opened the MP4 to MP3, MP4 to WAV and MP4 to AAC pages on a local development build, chose the sample, pressed Convert and then Download, using headless Chromium. All three downloaded and played back in ffprobe with the values in the table. The browser MP3 and WAV matched what ffmpeg 8.1.2 on the command line produced for the same settings apart from a few hundred bytes of MP3 header data (161,270 bytes in the browser, 161,479 with the command line). We tested only this small file and only the Chromium engine, so we do not claim the same speed for a 100 MB file or for Safari.

Ffprobe 8.1.2, our run on October 21, 2026.

Why the MP3 is 10.057 s and the WAV 10.008 s: MP3 and AAC encoders add a few thousandths of a second of padding. Matching numbers within a few percent mean the converter worked; a result of 0 bytes, a different duration by seconds, or a silent file means something is wrong with the input (see the silent audio section).

Why is my extracted audio silent?

We reproduced the common causes with ffmpeg on October 21, 2026.

What we saw and what to do.

Also check that the file plays with sound in a normal player first: an extraction cannot make sound the video does not have. The browser tool takes the default (first) audio track, so for a second track use ffmpeg.

Why is the audio shorter than the video?

When the sound starts after the picture, the extracted file begins at the first sound and the silent lead-in is gone. We built a 30 second video whose audio starts 1.976 seconds late (1,976 ms). The extracted MP3 and WAV were both 28.03 seconds long, 1.97 seconds shorter than the video, so laid back under the original picture the sound would play early.

To keep the lead-in, add the delay back when extracting. With the same file, ffmpeg -i video.mp4 -vn -af adelay=1976:all=1 audio.wav gave 30.00 seconds. The number is the delay in milliseconds; read yours from ffprobe (the audio stream start_time times 1000).

Small differences, a few hundredths of a second, are normal: MP3 adds a short frame of padding (our MP3 read 30.000181 s for a 30 s source).

Can I extract audio from someone else's video or from a link?

Only if you have the right to use it. Koldflux converters work on files you hold; they do not download videos from YouTube, Instagram or other sites, and we do not cover that. Extracting audio does not give you rights to reuse it, so check who owns the sound before you post it.

Limits and how we checked

• Tested by us: one generated 30 second 640 by 360 test video, audio extracted with ffmpeg 8.1.2 on October 21, 2026, with bytes and bitrate read by ffprobe and levels by the volumedetect filter. The 10 second browser results, the command line comparison and the level measurements are in the published test log.
• Not logged: the 30 second size table, the late-start figures and the 30.000181 s MP3 length come from an earlier session whose log was not kept.
• The browser converter was run end to end with headless Chromium on a local development build, with the page security policy relaxed because the development build needs it. A production build does not.
• Not tested: the QuickTime Player and Clipchamp steps come from Apple and Microsoft pages. Real files with other codecs or lengths may behave differently from our test file.