Superior pitch shifting through frequency domain?

[Dog Boots]

I've been wondering for several years:

We all know how pitch shifting algorithms are still a very lossy process that often creates very audible artifacts.

If compression schemes like MP3, OGG etc. convert everything to the frequency domain and back again at playback - wouldn't it be possible to manipulate the audio (simple multiplication?) while in the frequency domain and then get it back pitch shifted with only the quality loss of the compression which can be virtually non-existant depending on the bitrate.

Why is this not the case (since it hasn't been done a long time ago)?

[Philbo King]

Lots of media player apps already do this

[Dog Boots]

Originally Posted by Philbo King

Lots of media player apps already do this

Thanks for replying!
Really? Do you know a specific one that does it - just tried on VLC and Windows Media Player (the latter I can only make time stretch, but same deal....) and they both sound like they are doing the usual cutting things up and stitching them back together.

[DVDdoug]

The normal way to change pitch & speed/tempo independently is in the frequency domain. I think the main reason it's imperfect is because a batch of samples (the FFT length) gets converted to the frequency domain.

If you have an infinitely-long continuous signal then the FFT & reverse FFT can be perfect but of course real world audio signals change moment-to-moment.