You have a recorded call, a lecture or an interview as a video file, and what you actually want is the sound. Maybe for a podcast cut, maybe to transcribe it, maybe because a 900 MB video of a static slide is absurd when the content is a voice.
The audio is already a separate stream inside the container, so getting it out is more like unpacking than converting, as long as you ask for the right output. It is worth knowing this before reaching for an online converter, because uploading a file to one hands over a recording that a copy operation could have produced locally in seconds.
Copy or convert
The audio in an MP4 is almost always AAC. That gives you two very different operations:
Copy (remux). Write the existing AAC stream into an .m4a or .aac file. Nothing is decoded, nothing is re-compressed, the result is bit-for-bit the audio that was in the video. It takes about as long as copying a file.
Convert (re-encode). Decode the AAC and encode it as something else: MP3, Opus, WAV. This is a second lossy generation if the target is lossy, and takes real time.
Asking for MP3 out of an MP4 is therefore always a re-encode, even though MP3 feels like the "normal" audio format. If nothing downstream specifically demands MP3, taking the M4A is both faster and better.
WAV does not undo anything
A persistent belief is that converting to WAV recovers quality, because WAV is uncompressed.
It does not. Lossy compression discards information at encode time and there is nothing to bring back. Converting a 128 kbps MP3 to WAV gives you a file roughly ten times the size containing exactly the same damaged audio, stored without further compression.
WAV is useful for two real reasons: it is what editing and processing tools want to work with, because decoding on every operation is wasteful, and it avoids adding another lossy generation when you are going to process the audio repeatedly. Both are about the work ahead, not about the audio behind.
Choosing a bit rate for speech
Speech is much easier to compress than music, and most extracted audio is speech.
- 64 kbps mono AAC or Opus: clean for a single voice; an hour is about 28 MB.
- 96 kbps mono: comfortable headroom, still small.
- 128 kbps stereo: the default most tools use, and usually wasteful for a recorded call.
- 192 kbps+ stereo: for music. Nothing about a meeting recording needs this.
Two people talking over a conference call are already mixed into a single channel of information even when the file says stereo, so mono usually halves the size with no loss of anything you would notice.
Trimming before converting
An hour-long recording of which twelve minutes matter is an hour-long encode. Cutting the section you want first and converting the cut is faster, smaller and avoids a pile of audio you will never listen to.
This matters especially for transcription, where processing time scales with duration. Extracting, trimming and then running speech-to-text on the relevant twelve minutes is a fraction of the work of transcribing the whole thing and searching the result.
Removing audio instead
The opposite job comes up as often: a clip that needs to be shared without the room's conversation on it, or a screen recording where the microphone caught something it should not have.
Removing the audio track is a container-level operation on the video side: the video stream is copied untouched and the audio stream is simply not written. It is instant and does not degrade the picture.
That is also the safe way to handle a recording with a sensitive few seconds: strip the audio entirely, rather than trusting that nobody turns the volume up.
A short recipe
- Need the sound as-is → extract as M4A with stream copy.
- Need MP3 because something demands it → extract, accept one re-encode, 96 kbps mono for speech.
- Going to edit it → extract to WAV, work, export once to a lossy format at the end.
- Going to transcribe it → trim to the relevant part, extract mono at 64–96 kbps, transcribe.
- Need the video without sound → strip the track, do not re-encode the video.
MP4 to MP3 covers recipes 2 and 4: it takes the soundtrack from an MP4, MOV or WebM file, lets you choose the bit rate and mix down to mono, and works on the local file in the tab. It exports the whole soundtrack, so trim in an audio editor afterwards if you only need a section. For recipe 3, the audio converter writes WAV from a video's soundtrack too. Keeping this local is the difference between preparing an internal recording for transcription and uploading an internal recording to a third party to find out what is in it.