Two files out of one, with nothing thrown away
The [msm_tool slug="video-detach-audio"] splits a video into its two halves and keeps both: the picture as a silent video, and the sound as an MP3. That is the difference between this and muting. Muting produces one file and discards the audio; detaching produces two and discards nothing, which matters when the soundtrack is the part you actually intend to work on.
An MP4 keeps its picture untouched
Where the source is an MP4, the video stream is copied across exactly as it was, without being decoded and encoded again. There is no generation loss at all, and the job takes seconds rather than minutes, because copying a stream is cheap in a way that re-encoding never is. The silent video that comes out is, frame for frame, the video that went in.
Other containers cannot be copied that way in a browser, so their picture has to be re-encoded, which is slower and costs a little quality. The tool says which route it is taking before it starts, so the trade-off is visible in advance rather than discovered afterwards in the result.
The timing survives, so the halves go back together
Because the picture keeps its original timing, the two files line up again if you put them back into an editor and start both at zero. Nothing is trimmed, padded or resampled to make the split work. That is what makes the tool useful as the first step of an edit rather than a destructive operation: send the MP3 to whoever is cleaning up the audio, keep the silent video untouched, and recombine later without hunting for sync.
What people actually use it for
The common case is fixing sound that is worse than the footage. A phone recording of a talk, a walk-and-talk filmed outdoors, or an interview in a room with air conditioning all produce usable video and audio that wants levelling, noise reduction or a compressor — work that belongs in a sound editor, not a video editor. Detaching gives that editor exactly what it needs and leaves the picture out of harm's way.
The second case is reuse. A clip that needs a new soundtrack starts as a silent video, and detaching produces one without destroying the original audio in the process, so the old track is still there if the new one does not work out. The third is archival: the MP3 of a lecture or a panel plays anywhere a podcast does, on a commute or in a car, while the video stays available for the moment someone needs to see a slide.
The quality setting applies to the MP3 only
The bitrate choice affects the audio file and nothing else. It does not change the video, which is copied untouched from an MP4 regardless of which setting is selected. As elsewhere, 128 kbps suits speech, 192 suits music, and 320 only helps when the source audio was good to start with — a higher setting cannot add detail the recording never captured.
When there is nothing to separate
A video with no audio track cannot be split, because one of the two halves does not exist. Screen recordings made without system audio are the usual case. The file is checked first, so this is reported plainly rather than producing an empty MP3 alongside a copy of the video.
Long recordings and memory
The soundtrack is decoded into memory in full before it is encoded as an MP3, which is the same constraint the audio extractor works under. On a computer this is rarely a problem. On a phone, a recording over roughly half an hour can exhaust the memory a browser tab is given, and the same file will usually go through on a desktop without trouble.
Nothing is uploaded
Both files are produced by your own browser, and the video never leaves your device. For recordings of meetings, medical appointments, classrooms or anything involving people who did not consent to a third party holding a copy, that is the property that matters most — and it is why the work runs on your machine even though a server would be faster.