Pitch and Time shifting

[fl]

I was looking at the Prosonic MPEX plugin for Pyramix and went searching for information on the web. It appears that they are abandoning the Windows platform, and discontinuing support for their Windows programs after June of this year.

In light of this, what are some alternatives for Pitch and Time (duration) manipulation in/with Pyramix?

There are those situations where an unaccompanied singer or chorus drifts away from the reference pitch as the piece progresses. What's worse, inserts are often recorded starting in the middle of the piece, but with the "correct" pitch, which won't match up with what is in the master take at that point. I'm thinking that a process that applies a gradual shift from beginning to end of a clip would allow me to match up pitch at edit points more easily than going through the whole process of resampling whole takes and then editing.

My experiences with Melodyne Editor have been mixed - it's not so great when dealing with polyphonic material recorded in a reverberant environment. And anyway, I'm not at all sure that tweaking each individual note in a performance is what I'm after.

Pyramix's Resampler (in the Media Manager) does a passable job at changing an entire file, but the resampling also creates changes in tempo and duration. Just about any process that applies this kind of SRC (to or from a non-standard sample rate, i.e., other than 44.1, 48, 88.2, etc.) is going to introduce more artefacts than shifting from, for example, 96 kHz. to 44.1, but the program or plug-in I'm hoping to find would have a minimum of this interference/distortion.

What I'd really like is something like Sequoia's "Elastic Audio", which allows a gradual and increasing pitch shift over the duration of a clip (or in Sequoianese, an "Object").

I'm hoping to find something that would work either as a component of Pyramix itself, or as a plug-in that could be used with the Render process, or failing that, as a stand-alone program.

Thanks for any suggestions.

[Perfect Record]

It's slow and tedious, but the results are sonically pretty transparent. We've been using sample frequency conversion for pitch shifting of acapella music. The tempo changes are trivial for any sane pitch change. I calculate the desired pitch change and convert it to sample frequency, then do a family of pitch shifts (sfc shifts) around the target value, then do a couple test edits to chose the one that works best.

Yeah, I know there are faster tools, but this method is sonically clean.

[fl]

It's slow and tedious, but the results are sonically pretty transparent...

Yes, that's exactly what I have been doing as well, but the drawback is that you wind up shifting large sections of a piece of music. Once you shift one take, you find that the next one needs to be shifted to match, and so on... That Elastic Audio thing in Sequoia is pretty darn slick, and my hope is that it, or something similar, would allow me to wind up having to apply pitch correction to less material overall, both in the current take and all those subsequent.

I downloaded a demo version of Prosoniq's "TimeFactory 2", which does the SRC and keeps the duration the same (or changes the duration without changing the pitch). However, version 2 only works with complete files - no gradual shift. When I asked them about it, they state that it is something they're working on for version 3, but there is no release date for that yet - probably next year or later. It does a reasonable job of pitch shifting, with a choice of five algorithms, but for $475, I don't believe it's worth it for what it does compared to what Pyramix already does. As you mention, for shifts of less than a semi-tone, the tempo shifts are barely detectable, although I have noticed some slurring of certain consonants.

If you have a means of displaying a spectrograph of time vs. frequency, such as what is available in Izotope RX, and would like a chuckle/horrifying moment, take a look at what various methods of SRC do to the spectral content of a file consisting of a swept tone from 20 Hz. to 20 kHz. For the common conversions, from one "standard" SR to another, most of what's out there is pretty good (a more complete look can be found at http://src.infinitewave.ca ). The new Apodizing SRC available in Pyramix 7 is very clean and only displays a little aliasing resulting from frequencies above 22.05 kHz. (when converting a 96 kHz. file to 44.1). Other good converters include Izotope RX, Voxengo's r8brain, Sox, and Sequoia's Ultra High Quality 1 and 2.

It's a different story when you are only shifting pitch less than a semi-tone. Pyramix displays a lot of spurious frequencies, seemingly parallel to, or the inverse of, the original frequency sweep. In addition, there are weird "windows" whereby different frequency ranges suddenly generate a wholly different amount of extra components. On the spectrograph of the sweep, this shows up as blocks of noises that appear at certain points in time, and then give way to something else later on at a different frequency range.

From what I've seen, there is nothing that doesn't add noise and distortion for these small changes in SR - Sequoia adds a lot of alias frequencies that appear to be reflections of the original frequency sweep, as does the Prosoniq TimeFactory.

I haven't yet run a swept tone through the Sequoia Elastic Audio that does a gradual shift of pitch adjustment - that will be interesting. It' is my hope that the use of a gradual shift methodology, I'd wind up having to convert fewer takes, and thereby not have the bulk of an edited piece being comprised of shifted material. Of course, if they could sing in tune... (what a silly notion!)

But, if it's out of tune, it's out of tune, and it's got to be fixed if you're editing from one take to another - so, in the immortal words of Bob Ohlson, "Signal processing is always a trade-off between degradation and a perceived improvement in sound."