Auto detection editing of "Um" and "aah"
I do a lot of editing of interviews and many of the people end up having very disjointed speech styles with many pauses, "um's" and "aahs". Now sometimes I keep these in where it's natural, but sometimes you need to condense a section or improve it for clarity, so it has to get cleaned up.
With AI and machine learning becoming more of a thing recently in widespread use, is there room for an automatic detection feature where the audio is analysed and it can automatically generate cuts to remove ums and aahs etc? I assume that this would need manual cleaning up, but if you look at the kinds of "magic" editing being made possible by the audio editor Izotope RX, this kind of thing could be quite do-able, as it's easily possible to pick up certain repetitive types of sounds with predictable frequencies etc.
Obviously you'd need to cover over the jump cuts with b-roll footage. The goal is to get the audio content nice and clean and natural, without unnecessary bits, and then to cover over all of that with cutaways.
Perhaps there could be an option to have a morph cut visual transition added if desired, if you're only taking out a few. I find that morph cut only really works at very short durations in this kind of scenario, and only if there is minimal movement in the face. Approx 4 frames long on average.
Similarly you'd need a short cross fade audio transition - once again the AI or ML could come in handy here to insert the crossfade with the correct duration so that it's not letting a fraction of audio through on the wrong side of the transition that would make it messy. This feature could be very useful
Jon C commented
Seems like a no-brainer to have this feature added!
Ahmed Mostafa commented
I would love to
Daniel Schultheis commented