Convert an MP4 to MP3 without ruining the sound
Which bitrate to choose, why an MP3 pulled from a video always sounds quiet, and how to avoid compressing twice.
A 90-minute filmed conference, a tutorial where only the narration matters, a gig recorded on a phone: in all three cases the file runs to several hundred megabytes, when the sound on its own would fit into a tenth of that space. The job is to get it out of the container without mangling it on the way.
The operation takes thirty seconds, but two or three decisions made at conversion time settle how good the result will be. Here they are, along with what happens when things go wrong.
An MP4 doesn’t hold ‘a video’, it holds tracks
MP4 is a container, not a compression format. Inside it sit at least one video track (H.264 most of the time, H.265 on recent iPhones, AV1 on some platforms) and one audio track, almost always AAC-LC. Subtitles, chapters and additional audio tracks sometimes turn up as well.
Converting an MP4 to MP3 therefore means throwing away the video track and re-encoding the audio track from AAC to MP3. That detail has a direct consequence: this is not a simple extraction but a transcode between two lossy formats. The decoder rebuilds a signal from the AAC, then the MP3 encoder compresses it all over again and discards a second round of whatever it judges inaudible. This is generation loss: mild if the bitrate is chosen sensibly, plainly audible if it isn’t — cymbals turning into a hiss, a slightly metallic edge on voices.
For a sense of scale: a ten-minute 1080p video commonly weighs anywhere between 200 MB and 1 GB depending on the encoding bitrate, while its AAC audio track at 128 kbps accounts for under 10 MB. All the weight is in the picture.
Bitrate, the one setting that really matters
The bitrate sets how many kilobits are allocated to each second of sound. It governs both the file size and the amount of damage. Rough sizes per minute of stereo audio:
- 320 kbps — around 2.4 MB per minute. The MP3 ceiling. Worth it only if the source is good quality.
- 192 kbps — around 1.4 MB per minute. The best compromise for music taken from a video.
- 128 kbps — around 1 MB per minute. Enough for speech, fine for background music.
- 64 kbps in mono — around 0.5 MB per minute, so roughly 29 MB for an hour-long interview. Ideal for a meeting recording or a lecture.
The most useful rule is counter-intuitive: pushing the bitrate above that of the source restores nothing. If the original AAC track is 128 kbps, a 320 kbps MP3 will merely store an already-truncated signal in finer detail — two and a half times the size for no gain at all. Dropping well below the source bitrate, on the other hand, stacks two rounds of destruction, and that you can hear. Aiming for an MP3 bitrate close to, or slightly above, the source is the right instinct.
Two other settings deserve a glance. The sample rate of video is almost always 48 kHz, whereas catalogue music sits at 44.1 kHz. MP3 handles both, so there is no point resampling: it is one more processing stage for nothing. And mono lets you halve the bitrate, and therefore the file size, with no perceptible loss on a spoken voice — the channel count alone doesn’t determine the size, the bitrate does. There is one exception worth knowing about, though: if the recording has a microphone on one channel only, the downmix actually evens things out.
How to do it
Convertu’s MP4 to MP3 converter accepts MP4, MOV, AVI and MKV files. Drop the file into the upload area, start the conversion, collect the MP3. Processing happens on our servers — unavoidable for video decoding — and uploaded files are deleted after an hour.
One habit that saves time later: rename the file before you upload it. The resulting MP3 inherits the source filename, and since a track pulled out of a video carries no ID3 tags, that name will be the only information your audio player has to go on. A file called VID_20250412_143201.mp4 gives you an MP3 that shows up as ‘Unknown Artist’ in a library of several hundred tracks.
Where it goes wrong
Several audio tracks
A file ripped from a DVD, a recording of a bilingual conference or a multitrack edit can carry two, three or five audio tracks. An automatic conversion keeps a single one, usually whichever has the most channels — not necessarily the one you want. If you end up with the dubbed track when you were after the original-language soundtrack, open the file in a player that can list the tracks (VLC does it under the Audio menu) to work out which is which, then export that track from an editing package before converting.
5.1 multichannel audio
MKV or MP4 files that come from commercial discs or a TV recorder — DVD and Blu-ray rips, set-top box captures — often carry AC-3 or E-AC-3 in 5.1. MP3 does not go beyond two channels in practice, so moving from 5.1 to stereo forces a downmix. The centre channel, which carries most of the dialogue, gets folded in with the front channels: the classic outcome is a soundtrack in which the music swamps the voices. That isn’t a converter bug but an unavoidable consequence of the fold-down: the centre channel is attenuated as it is mixed in, whereas in 5.1 it was isolated and perfectly intelligible.
The file is rejected, or the conversion fails
Three causes come up again and again. First, the mislabelled extension: a file renamed to .mp4 when it is really a .flv or a .webm. The converter reads the actual header rather than the name, and turns it away. Running it through Convert Video Format to produce a genuine MP4 sorts that out. Then incomplete files: an interrupted download leaves an MP4 whose index (the moov atom) is missing or truncated, and nothing can be decoded. Finally, DRM-protected content, bought or rented from a platform: the audio track is encrypted and no web tool will read it. This is also the moment to point out that extracting the soundtrack from a work you don’t hold the rights to doesn’t become lawful just because it happens to be technically easy.
The file is too big to upload
An upload size limit applies, and it varies from one tool to the next: the figure that counts is the one shown on the tool page, under the upload area. Shrinking the video first with a video compression tool gets you under the threshold, but take care: if you lower the audio bitrate at the same time, you are adding another generation of compression. Check that the compression is being applied to the picture. The other approach, often cleaner, is to cut the video into segments in a desktop application and convert each segment separately.
MP3 isn’t always the right choice
MP3’s advantage is that everything plays it, car stereos and old MP3 players included. Depending on what you need, though, another format may serve you better.
- For editing or mastering, go for WAV. It is bulky (roughly 10 MB per minute at 16-bit stereo) but adds no further loss to the already-decoded signal. The Extract Audio tool offers WAV and AAC output alongside MP3.
- To stay as close to the source as possible, AAC or M4A output makes more sense: it avoids hopping from one lossy codec family to another, which limits the damage. It doesn’t cancel the damage out, though — only copying the track without re-encoding would be strictly lossless.
- If your end goal is text — meeting minutes, lecture notes, subtitles — you don’t need a good-quality MP3: mono at 64 kbps is more than enough for a speech recognition engine, and the lighter file will import faster into whichever transcription tool you use.
Two checks after converting
The first concerns loudness. Video is normalised to around -23 LUFS for European television broadcast, or -14 LUFS for streaming platforms, whereas commercial music is mixed a great deal louder. An MP3 taken from a video will therefore always sound quiet dropped into the middle of a playlist. That isn’t a conversion fault: the file is faithful to its source. A pass through a loudness normaliser closes the gap if you need it for listening on the go.
The second concerns metadata. Add title, artist and year in the file properties once you have the MP3, or it will end up orphaned in your library. On Windows, a right-click then Properties / Details is enough; on macOS, the Get Info pane in Music does the same job.
Video conversions need a machine that can decode, so the work happens on the server: one conversion a day is free without signing up, and unlimited access to the whole catalogue comes with a single subscription at €7 a year, cancellable in one click.