I change speed, and then change pitch,… and then its quality gets worse significantly.
This doesn;t seem to happen with other operations.
Only change pitch has very negative effect. Change speed doesn’t hurt much.
I partially understand why this happens: because change pitch is not a natural operation.
By natural, I mean, something like linear. A pure stretch is a simple operation, but stretching each wavelength
separately is a tricky thing, and is not natural.
Is there a fix?
I mean, something like linear.
Right. Change speed works like dragging your finger on the record. It’s not perfect, but it creates minor theatrical errors, not content errors.
Change Duration and Change Pitch have to take sound sound apart and put it back together—oh, and finish in the same day you started. If you’re going to have problems, those two are good candidates.
None of these tools will survive large changes. If you’re that far off, you probably need to either reshoot or get different content.
Many people run into the Change Pitch problem in the “Make Me Into An Announcer” job. That’s actually two problems. The actual pitch change is ratty and not all parts of the voice changes between people. You can’t tell people apart when whispering, but the instant they start speaking, the Identification is immediate. Change Pitch changes everything.
Maybe using echo might fix this problem a little bit.
A sound is heard now and a quarter second later. (not sure if this is what Echo does).
I may try it.
I think only possible real solution would be some high tech AI treatment.