Audio
Extracting the audio from a video: MP3, M4A, WAV or FLAC, what each choice weighs
Keeping only the soundtrack of a video is a common need: listening to a talk again on the commute, recovering the music from a filmed show, sending a voice message recorded as a video, preparing an excerpt for an edit. The picture is almost always the heavy part of the file. We extracted the sound from a real smartphone video in each of the formats FileXvert offers, with its exact settings, to see what each choice weighs and what it keeps.
Sound is only a small part of a video
Our test file is a video shot on a smartphone in 2015: 88.75 seconds in 1080p, in MOV format. It weighs 261.3 MB. Its picture is encoded in H.264 at about 22 megabits per second; its sound is stored with no compression at all (PCM, 16-bit, 44.1 kHz, stereo), that is 1.4 megabits per second. Even uncompressed, the sound therefore accounts for only about 6% of the file.
Extracting the audio means keeping only that track, then saving it in the chosen audio format. FileXvert tells FFmpeg to ignore the picture (the -vn option), then encodes the sound with the encoder matching the requested format. Everything happens in your browser: the video is not sent anywhere, which matters when it is a meeting recording or a family moment.
What we measured
We applied to this video the commands FileXvert runs for each output format. For lossy formats, the bitrate depends on the compression level chosen: 192 kb/s with no reduction, 128 kb/s in “balanced”, 96 kb/s in “smallest size”.
| Output format | Setting | Size | Share of the video |
|---|---|---|---|
| WAV | 16-bit PCM, lossless | 15.6 MB | 6.0% |
| FLAC | lossless | 2.54 MB | 0.97% |
| MP3 | 192 kb/s (no reduction) | 2.13 MB | 0.82% |
| MP3 | 128 kb/s (balanced) | 1.42 MB | 0.54% |
| MP3 | 96 kb/s (smallest size) | 1.07 MB | 0.41% |
| M4A (AAC) | 192 kb/s | 2.14 MB | 0.82% |
| M4A (AAC) | 128 kb/s | 1.44 MB | 0.55% |
| Raw AAC (.aac) | 192 kb/s | 2.15 MB | 0.82% |
| OGG Vorbis | 192 kb/s target | 1.37 MB | 0.53% |
MP3 at 192 kb/s divides the size by more than 120 compared with the video, and by more than 7 compared with the uncompressed sound. The WAV has exactly the expected size: 1,411 kb/s for 88.75 seconds, the original sound copied as it is.
The surprising case of FLAC
FLAC is a lossless format: the decoded sound is identical, sample for sample, to the WAV. Here, though, it weighs only 16% of the WAV, barely more than a 192 kb/s MP3. The explanation lies in the content: our video records a quiet ambience, with an average level of -49 dB. A faint, steady sound is easy to predict, and FLAC compresses precisely by predicting each sample from the previous ones.
This result does not generalise. On a recording of dense rain, measured in our guide to MP3 bitrates, FLAC still keeps 62% of the WAV's size. For rich music or a noisy crowd, expect a FLAC of around half the WAV, and therefore much heavier than an MP3.
Which format to choose
The right format depends on what you will do with the sound:
- To listen on a phone, a speaker or in the car: MP3. It is the most universally supported format. 128 kb/s is enough for speech; keep 192 kb/s for music.
- For an Apple device, iTunes or editing on an iPhone: M4A. The sound is AAC, the same size as MP3, in an MP4 container that handles metadata well. Prefer it to raw .aac, which players recognise less reliably.
- To rework the sound in audio editing software: WAV or FLAC. They add no loss, which matters if the sound will be encoded again later.
- To archive losslessly while saving space: FLAC, played by most modern players, but not by every older device.
- For a web page or open-source software: OGG Vorbis works well, but remains less compatible than MP3 with consumer devices.
Three precautions
First, the output format cannot improve the original sound. If the video already contains compressed sound, such as the AAC in most recent smartphone videos, extracting it as WAV or FLAC gives a large file without gaining any quality: it faithfully preserves a sound that has already lost some of its detail.
Second, an MKV video can contain several audio tracks, for example several languages. FileXvert extracts only one: the one FFmpeg picks by default, usually the track with the most channels. If you need another language, check the result.
Finally, extraction converts the whole length of the video. To keep only one passage, trim the video before converting, or the sound in an audio editor afterwards.