Recording a YouTube video’s sound

A YouTube video plays inside a frame of YouTube’s own, and no script on the page can read its sound; there is no file of it anywhere the toolbox can reach. What Chrome and Edge on a computer do allow, once you say so, is a recording of the tab itself, sound and all. So for a YouTube video the cut editor first plays the stretch of the video once and records the tab while it plays, and then cuts the clip out of that recording, in the browser.

1. Where it works#

Anywhere else 🔊 cut the audio… is greyed out, and the note under it says why:

The page asks for a browser built on Chromium, as Chrome and Edge are, so another one built on it passes too.

2. Recording, step by step#

1. cut the audio… opens the editor on a sentence saying what is about to happen — A YouTube video’s sound is cut from a recording of this tab: the stretch 11.13–15.40 s plays once, sound on, while it is recorded. Chrome asks first: press “Allow”, and leave “Also allow tab audio” turned on. — and a ● record button, which Enter presses too.
2. Press it. Chrome asks whether to share this tab, with Also allow tab audio switched on. Leave it on, and press Allow. While the tab is shared, Chrome says so in a bar across the top of the page, with a Stop sharing button, and the page moves down a little to make room for it.
3. The video plays the stretch the strip will show — the sentence and a second either side — from half a second before it to half a second after, sound on, at full volume and at normal speed, while a bar in the editor fills and the status line counts the seconds: recording 11.13–15.40 s of the video: 2.10 of 4.27 s. The toolbox’s own players on the page are silenced meanwhile. Afterwards the video goes back to where it stood, with its own mute, volume and speed.
4. Then the editor works on that recording as it would on a narration: recorded 10.63–15.90 s: move the edges by ear, then save the clip. The strip, the rows, the keys and the previews are the ones described in Cutting the audio; the previews play the recording, not the video, and the waveform is drawn in the browser from the recording itself, ffmpeg or not.
5. save clip cuts the clip out of the recording there and then, as a WAV with 8 ms of fade at each end, and sends it to the clip tray, which keeps it as an MP3 (or the best format the server’s ffmpeg writes) when the server has ffmpeg, and as the WAV when it has not.

A pass in which the video stopped to load, or whose timing did not hold together, is played once more, and the status line says so — (once more: the first time did not hold together).

One Allow lasts while the page stays open. Every later clip, and every frame captured for a card (A frame of the video), uses the same share, and Chrome does not ask again: with the tab already shared, sound and all, the recording starts as soon as the editor opens, and its sentence no longer speaks of Chrome. A share given without its sound (Also allow tab audio turned off for a frame, say) still takes frames, but the next cut lets it go and asks again, and the share given then takes the frames too. Stop sharing, or reload the page, and Chrome asks the next time.

3. The recording holds only what was played#

You may move an edge beyond what was recorded — the strip shows the stretch — but a clip that reaches past the recording is not saved: the recording holds 10.63–15.90 s: record again to take in the edges. A preview there says that is outside the recording: record again to hear it. ● record again records the stretch the strip shows now, a second past the moved edge.

When the caption is not where the sentence is#

YouTube’s transcript cuts a sentence across often enough: the words you want begin a second or two before the caption does, or run on after it, and the strip has nothing there to drag an edge into. The reach row under the two edges says how much further ● record again should go — seconds typed into its two boxes, start and end, negative earlier and positive later, up to a minute either way. Enter in either box records at once.

What comes back is a longer recording, and the strip grows to show all of it, so the edges reach into what the caption left out. The numbers are measured from the stretch itself, not from the last recording, so pressing ● record again twice with the same numbers goes to the same place rather than creeping further out each time.

A clip saved before a new recording is dimmed, and the status line says the clip saved is of the recording before this one: “use this clip” cuts it again, from this one; save clip or use this clip cuts it again from the new recording, and the old clip leaves the tray.

4. What the editor may say#