Extract Audio from Video Online

It takes the soundtrack out of a video and hands it back as an audio file. The picture is discarded, never decoded, and never re-encoded — so a two-hour recording costs roughly what its audio track costs, not what the video does.

• Lossless when it can be: if the video's audio is already in the format you asked for, the encoded audio is copied out untouched. The result is identical to the original track, sample for sample.
• Six output formats: MP3 and AAC/M4A for sharing, OGG for the web, WAV and FLAC when you need every sample intact.
• Nine input containers: MP4, MOV, WebM, MKV and OGV, plus AVI, FLV, WMV and 3GP — formats most video tools refuse outright.
• Trim before you extract: pull four minutes out of a two-hour lecture without waiting for the other hundred and sixteen.
• Loudness and levels: even out a recording whose volume wanders, hit a podcast or broadcast loudness target, or nudge the whole thing up or down.
• Multi-track videos: choose which audio track to take when a file carries a second language or a commentary.
• Nothing is uploaded: your video is read on your device, not our servers. There is no queue and no server copy.

Audio inside a video is already encoded. When the format you ask for is the same one the video is carrying, there is nothing to convert: the encoded audio can be written straight into a new file. No decode, no encode, no quality loss, and it finishes almost instantly.

Most online extractors re-encode every file regardless, which quietly costs you a generation of quality on audio that was already the right format. This tool detects the case and takes it, and the panel tells you before you press anything which of the two you are about to get.

When a copy is possible

• AAC audio (nearly every MP4 and MOV) into M4A or AAC.
• Opus audio (nearly every WebM) into OGG.
• Vorbis audio (OGV and some WebM files) into OGG.
• MP3 audio (common in AVI and FLV) into MP3.
• FLAC audio into FLAC, and uncompressed audio into WAV.

When it is not

Anything that changes a sample turns the copy back into a conversion: picking a bitrate, changing the sample rate, mixing to mono, normalising loudness or adjusting volume. That is not a limitation so much as arithmetic — you cannot both leave the samples alone and change them. The tool never applies one of those settings silently on the copy path; if you ask for one, it says the extraction will be re-encoded and why.

The honest answer is that the audio in your video already has a quality ceiling, and no output format can raise it. What the format decides is size, compatibility and whether anything further is lost.

The six, and what each is for

• MP3: plays on everything ever made. Pick it when the file is going somewhere you do not control. 192 kbps is transparent for speech; 256–320 kbps for music.
• M4A (AAC): better than MP3 at the same bitrate and the native format of every MP4. If your video is an MP4, this is usually the lossless copy — take it.
• AAC: the same audio as M4A in a bare stream rather than an MP4 wrapper. Useful for hardware and broadcast chains that expect it; M4A is the friendlier choice otherwise.
• OGG: the open-format choice, and the lossless copy for anything recorded as WebM. Well supported on the web and on Android, less so in older desktop software.
• WAV: raw, uncompressed, and the safest thing to hand to an editor or a transcription service. Large — roughly 10 MB per minute in stereo.
• FLAC: lossless like WAV but around half the size. The right archive format, and the right choice when the source audio was already lossless.

Extracting the whole of a two-hour lecture to keep four minutes of it is a waste of time and disk. Set a start and an end and only that span is written. Both boxes take plain seconds or a timestamp — 90, 1:30 and 0:01:30 all mean the same thing.

Trimming does not cost you the lossless path. On the copy route the cut lands on an audio frame boundary, which is a few tens of milliseconds either way — imperceptible, and the price of not re-encoding. When the extraction is being re-encoded anyway, the cut is exact to the sample.

The fields are bounded by the length of the file you actually loaded, so a start past the end of the recording is refused before anything runs rather than producing an empty file.

A recording made on a phone across a room is usually quiet and uneven. Three normalising modes cover the cases that come up, and all three re-encode by definition.

The three modes

• Smart: evens out a recording whose level wanders — someone leaning towards the microphone and then away from it. It adapts continuously rather than applying one fixed gain.
• Podcast (-14 LUFS): the integrated loudness target Spotify, Apple Podcasts and YouTube normalise to. Use it when the file is going to a platform.
• Broadcast (-16 LUFS): the streaming and broadcast target, a little quieter and with more headroom left for whatever processing comes after.

Alongside those, a volume control moves the whole file up or down by a fixed amount, and a channel control mixes a stereo recording down to mono — which halves the size and is usually the right call for speech recorded on a single microphone.

Most video tools on this site accept MP4, MOV, WebM and MKV, because they have to decode the picture and no browser can decode the codecs inside older containers. This tool never decodes a picture, so it can accept more.

What goes in

• MP4, MOV, WebM, MKV, OGV: read directly, with no size ceiling beyond your plan's. These are the containers where trimming and the lossless copy are fastest.
• AVI, FLV, WMV, 3GP: read by a second path that has to load the whole file at once, so these are capped at 150 MB. That is a real memory limit rather than a plan decision, and it applies whichever plan you are on.
• Anything else: refused at the file picker with the format named, rather than accepted and failed later. MPEG program streams (.mpg, .vob) are the notable one — convert those to MP4 first.

A video with no audio track at all is refused immediately with that reason, instead of producing a silent file that looks like it worked.

• Podcasts and interviews: pull the audio out of a recorded video call so it can be edited, levelled and published as audio.
• Lectures and meetings: keep the soundtrack of a two-hour recording as a 60 MB MP3 instead of a 4 GB video.
• Transcription: most transcription services want audio and charge by the file. WAV or a 128 kbps MP3 is what they ask for.
• Music and reference tracks: take a mix out of a live performance recording without re-encoding it.
• Video editing: extract a clean audio track to work on separately, then put it back with the replace-audio tool.
• Archiving: keep the part of a recording you will actually return to and drop the rest.

Free handles one video at a time, up to 300 MB and up to an hour long, and exports at up to 192 kbps and 44.1 kHz — which is a genuinely useful extractor, not a demo. The lossless copy path is free too, and deliberately so: it is the fastest and best-quality route the tool has, and putting it behind a payment would mean free users always got a worse file than the one already inside their video.

Pro raises the ceilings rather than unlocking features: up to twenty videos in one run with a single ZIP download, file sizes limited only by what your device can handle, up to 320 kbps, sample rates up to 192 kHz and 24-bit output for WAV and FLAC.

Your video is opened and read on your own machine. It is not uploaded, not queued on a server, and not stored anywhere — which is why there is no waiting line and no limit on how many times you can use it. It also means a recording you are not allowed to send to a third party can still be processed here.

Powered by optimised video encoding pipelines.

• Replace audio
• Video cutter
• Video compressor
• Video resizer
• Video cropper
• Video to GIF
• Video converter
• Audio converter
• Audio cutter
• Audio compressor
• Audio merger
• Blend audio tracks