Source—
Audio codec—
Output—
Re-encoded—
Almost every tool for this converts to MP3, and almost every time that is the wrong answer. The audio inside a video is already a finished, compressed track — usually AAC from a phone or a camera, Opus from a screen recorder — and a container can simply be rewritten around it. That makes extraction a copy: bit-identical to the track in the video, as fast as copying a file, and with none of the second lossy pass a conversion to MP3 would impose. This tool reads what is actually in the file, copies it out by default, and tells you what share of the video's size the sound was.
Step by step
01Open the video
MP4, MOV, WebM, or MKV. The tool reads the container and reports the codec, channel count, and sample rate of the audio track inside it.
02Take it as-is
The default writes the existing track into the container that matches its codec — AAC into an M4A, Opus into an OGG. No decoding, no encoding, nothing lost.
03Or convert, knowing what it costs
WAV if it is going into an editor, OGG for the web. Each of those decodes and re-encodes, and the ledger says so rather than leaving it implied.
04Take part of it, if that is all you need
The waveform's handles work here too — extracting and trimming are the same operation, so there is no reason to do them as two steps.
A worked extraction
The sample clip this tool opens with: ten seconds of 640×360 video with a mono Opus soundtrack, generated in the browser rather than downloaded. Every figure is what the tool reports on it.
Given
The same audio, three ways out
as-is (OGG) 31 KB copied 23% of the file
WAV 749 KB re-encoded 24× the OGG
MP3 — refused no browser can encode MP3
the picture was the other 77% of those 135 KBTwenty-three percent is high because the sample is a small, low-bitrate clip; on a real 1080p recording of a talk the audio is nearer a twentieth of the file, which is exactly why an audio-only copy is worth making — it streams on a connection the video would not. The important row is the first one. Copying the track costs nothing at all: no decode, no quantisation, no time. Converting it to WAV produces a file twenty-four times larger that contains no more information than the OGG did, and converting it to MP3 — the thing most tools do by default — would push already-lossy audio through a second lossy encoder, which is a real if small loss taken for no reason. It also cannot be done in a browser at all, because browsers ship an MP3 decoder and not an encoder.
Before you extract
A container is not a format
An MP4 is a box. Inside it are separate video and audio tracks, each in its own codec, and taking one out is a matter of writing a new box around it. This is why extraction can be instant and lossless, and why 'convert MP4 to MP3' is a misleading way to describe a job that mostly involves no conversion.
The audio is already compressed, so re-compressing it costs
AAC and Opus are lossy: they have already discarded what a listener will not miss. Running that through another lossy encoder discards more, and spends bits describing the first encoder's artefacts. It is a small loss, and it is entirely avoidable when the operation was only ever going to be a copy.
MP3 is a habit, not a requirement
MP3's ubiquity is thirty years old and every device that plays it also plays AAC. Taking an M4A out of a phone video plays everywhere the MP3 would have, is smaller at the same quality, and is bit-identical to what was in the video. Choosing MP3 out of habit costs quality and gains compatibility that was never in question.
Copying is bounded by disk, not by processing
A conversion has to decode and re-encode every sample, so it scales with the length of the recording — which is why upload-based tools queue. A copy is bounded by how fast the bytes can be read and written, so a two-hour recording takes about as long as saving a file.
No audio track looks exactly like a broken file
Screen recordings made without system audio, silent phone captures, and videos whose sound was already stripped all produce a file with no audio track at all — and a tool that fails vaguely on them sends you looking for a problem that does not exist. The correct response is to say the track is not there.
The recordings people extract from are the private ones
Interviews, recorded calls, lectures, footage that has not been released. Uploading a whole video to a free service to get the sound out of it is a poor trade for a job that a browser can do locally in the time it takes to read the file.
The judgement call
What you are going to do with the audio decides it.
Listening to it, or filing it
As-is
A copy of the original track: nothing lost, nothing to wait for, and it plays on everything that plays the video did.
Importing into an audio editor
WAV
Uncompressed, so the editor is not decoding on every operation. Large — 24× the compressed track on the sample here.
Putting it on a web page
OGG
Opus is the smallest of these at a given quality and every current browser reads it. Often this is also the as-is copy, if the video was a screen recording.
Sending to someone with an old device
As-is
AAC has been universal since the first iPod. The compatibility argument for MP3 has not been true for a long time.
You specifically need an MP3 file
Not possible in a browser
There is no MP3 encoder in any mainstream browser. A desktop tool is the honest answer, and this one says so instead of failing obscurely.
You only want a minute of a long recording
Drag the handles first
Extracting and trimming are the same operation here, so taking only what you need costs nothing extra.
Reference
FAQ
Copy the audio track rather than converting it. The sound inside a video is already a finished compressed track, so a tool can write a new container around it — the result is bit-identical to what was in the video. The Tinapps extract audio tool does this by default and reports the codec it found, so you can see that no conversion took place.
M4A, in almost every case. A phone or camera records AAC, which goes into an M4A container as a straight copy with nothing decoded; converting the same audio to MP3 pushes already-lossy data through a second lossy encoder for a compatibility benefit that has not been real for twenty years. It is also the only one a browser can produce, since no mainstream browser ships an MP3 encoder.
Yes. The Tinapps extract audio tool reads the video's container with your browser's own codecs through WebCodecs and writes the audio track out locally, so nothing is sent anywhere. That matters for the recordings people usually extract from — interviews, calls, and unreleased footage.
Because the picture is almost all of a video file. On the sample in the Tinapps extract audio tool the sound is 23% of a small low-bitrate clip; on a 1080p recording of a talk it is nearer a twentieth. That ratio is the reason to make an audio-only copy at all — it streams on a connection the video would not.
A few seconds, if the audio is copied rather than converted — a copy is bounded by how fast the file can be read from disk, so a 2-hour recording takes roughly as long as saving a file of that size, typically under 5 seconds. A conversion is different: it decodes and re-encodes every sample, so its time scales with the length of the recording, which is why upload-based services queue long files. The Tinapps extract audio tool copies by default and says which of the two happened.
Then there is nothing to extract, and the tool says so rather than failing vaguely. Screen recordings made without system audio, silent phone captures, and videos whose sound was already removed all produce a file with no audio track, which otherwise looks exactly like a broken file.
Yes — the waveform's handles work here, so extracting and trimming are one operation rather than two. Taking only the minute you need costs nothing extra and avoids writing out a two-hour file to cut it afterwards.
Not in the Tinapps extract audio tool, because nothing is uploaded and there is no bandwidth or storage to meter. The practical limit is the memory a browser tab can allocate while reading the container.
The video is opened and its audio track written out entirely in your browser. Nothing is uploaded, and the audio exists only as the file you download.
Keep going
MP3 cutter
Trim the extracted audio, with the cut snapped to a silence.
Video and audio tools
The rest of the category, all on the browser's own codecs.
MIME types
What audio/mp4, audio/ogg and audio/mpeg actually describe.
Word counter
For the transcript, once the audio is out of the file.