Sooner or later everybody finds that pitch change software never does what you want. Not only will it not change a Network Radio Presenter into a four-foot tall woman (we get that one every day) but it doesn’t even do relatively modest pitch changes without obvious damage.
I propose that many of these problems derive from the sound format being not up to the job. MP3 aside – nobody seriously expects that to work – I can put up a good argument for 44100 not being able to deal with sound accurately up past about 17 KHz. What happens if you take a 44100 sound track and pitch shift it down several tones? Suddenly that “potential” error up where only dogs can appreciate it, is down where even I can hear it and also, acoustic tricks and fancy-pants CD filters don’t help any more.
So, given that Audacity doesn’t appear to have any limits to the sound specification, what happens if you record a voice at 32-bit floating and something nose-bleed high like 96KHz or 128KHz sample frequency – solely to pitch shift it up and down and see if the damage is reduced, or eliminated?
I’m perfectly clear what happens when you push a whole range of sounds up and down, what I’m after is addressing the complaints of oddball high frequency frying and bubbling and thumping audio that weren’t apparent in the original show. I guess I’m after a perfect simulacrum of dragging my finger on the tape machine capstan. I have heard a guy performing reasonably well on tape that turned out to be a woman running at the wrong speed. We should be able to do that without a speed change.
I would have been happier with a human voice recorded live at various sample rates and bit depths. You can hide piano damage, but anyone can tell that there’s something wrong with a voice.
I wonder if there’s anything wrong with a Master Voice Clip at very high quality and downsampling to a lower rate for other tests. I instinctively hate that idea, but I can’t think of anything actually wrong with it.
Increasing the sample rate before applying Pitch shift appears to produce a slight improvement to the sound quality, but so slight as to be hardly worth the effort.
The “Sliding Time Scale/Pitch Shift” uses a different code base from the standard Audacity “Pitch Shift” effect and can produce a pitch shift effect that is very noticeably better quality than the standard Audacity “Pitch Shift” effect.
The “Sliding Time Scale/Pitch Shift” is very sparsely documented in the manual at present, but the effect is still being actively developed, so it’s probably not worth extending the documentation until we know the final form of the effect. Here’s what there is so far:
Sliding Time Scale / Pitch Shift
This effect allows you to to make continuous changes to the tempo and/or pitch of a selection. For example, you could set the initial tempo to -50% and the final tempo to +20%. Or you could set the initial pitch to +3 semitones and the final pitch to +1 semitones. Or you could do both at once.
To create the same (but in higher quality) effect as the standard Audacity “Pitch Shift” effect, simply set the initial and final pitch change to the same value, and leave the initial and final tempo change at the default (0.0 = no change).
Warning! The “Sliding Time Scale/Pitch Shift” effect in Audacity 1.3.12 is buggy, but it should be fixed in the next (1.3.13) release.
Koz, besides your post being absolutely hilarious, it really helped!
Just started using Audicity and almost tossed it out when I heard the hideous quality after doing the standard pitch shift. After following your advice, I swear the shifted track sounds better than the original! Fantash! Thank you.