Re: semi-ot: pitch shifters

[Per Boysen <perboysen@gmail.com>]

I think your questions are a bit too vague to be correctly answered,
even though Stefan gave a good Max/MSP introduction as usual ;-). For
example you are not telling us what you mean by your Eclipse "can't
keep up with the audio". It is also not clear what you mean by pitch
shifting not being "reliable and accurate". There might be issues with
playing chords into a pitch shifter... Also pitch shifting accuracy
depends on the frequency, if you are shifting a high or low not and if
you are transposing audio up or down in pitch. A qualified guess would
be that you are talking about live audio? (as opposed to audio being
recorded into RAM or elsewhere before pitch transformed).

I can't comment on Stefan's reply but I would like to point out that
instant pitch shifting is not physically possible in a real-time live
input signal. It is indeed possible with plugins that manipulate the
playback of an already recorded audio file on a computer's drive, but
then the software is working with preview calculation as the trick to
get to it right on time. In the real-time world a calculation has to
be performed a little bit "too late" for what is musically correct.
Maybe this is what you man by "not accurate"?

For pitching a signal without using the dry signal in the output I
have experienced some pitch shifting stomp boxes that feel very fast,
like the classic Blue Box by MXR, the Hog by Electroharmonix etc etc.
For a while I used a TC Electronics FireworX that did an "almost OK"
job on pitching down. Except for the FireworX those are analog boxes.
In computer software I doubt that would be possible, at least I have
not seen any software algorithm that transposes live audio without too
much latency. I'm not a programmer but I suspect all audio processing
programming has to rely on setting up buffers in the time line in
order to secure calculations being reliably performed. A computer's
operating system may also add a time buffer to the process. A while
back you could hear the difference as the same plug-ins under OS X
were "slower" according to the extra safety buffer in that system
compared to when run under Windows. That might have changed now
though, I haven't seen any whining about it lately ;-)

Myself I gave up pretty soon on finding a real-time pitch transformer
software. Instead I do as we musicians always do, we make a strong
gesture in some direction that (hopefully) makes the listener believer
they are hearing it right (although it actually is not perfectly
"right"). Such "cheats" include transforming already recorded slices
of a loop (Mobius scripting), using another physical instrument that
actually allows for playing the notes directly at the preferred pitch
and finally to blend the pitch transformed signal with the dry live
signal to mask out the ugly latency.  This last approach works fine
for me with pitch shifting software plug-ins. Ok, so here's sharing
some experiences with different pitch shifting plugins I have tried:
The least CPU intensive one I have found is the PItch Shifter II in
Logic/Mainstage but you have to set its quality to "vocal" for minimal
latency and fidelity degradation and it still sounds rather crappy if
listened to without the masking source audio signal. The cheapest is
Apple's AU Pitch Shifter that comes for free within OS X on any Apple
pc. I used this a lot when building a Mobius looping host in Bidule, a
little slower than the Logic plugin but still ok. Then I have also
tried the free plugin Spectral Magic AU suite by Michael Norris; it
has a better sounding pitch shifter but one that consumes a bit more
CPU power. I too am a big fan of Eventide's processing so I invested
in the SoundToys Native Bundle plugin suite as soon as it became
available as AU. For long time it was only offered as RTAS to be used
in Proo Tools but today there is even a Windows version of these
plugins. One of those plugins, the Crystallizer, offers very good
pitch shifting. Not as timing accurate for a realt-time live input as
an analog stomp box of course, but if you like the classic Eventide
harmonizers this is as close you get in software (and you get a lot
more routing experimentation options than the HM8000 ever had).

Best regards

Per Boysen


On Thu, Dec 9, 2010 at 2:02 AM, james fowler <twostroke@gmail.com> wrote:
> hey gang -
>
> below is a reply i got from stefan.  anybody else care to chime in?
>
> - jim
>
> ---------- Forwarded message ----------
> From: Stefan Tiedje <Stefan-Tiedje@gmx.de>
> Date: Wed, Dec 8, 2010 at 10:51 AM
> Subject: Re: max/msp vs. eventide
> To: james fowler <twostroke@gmail.com>
>
>
> Hi James,
>
> Am 08.12.10 15:04, schrieb james fowler:
>> i'm currently using an eventide eclipse and i'm finding that its
>> pitch shifters cannot keep up with the audio i'm feeding it.  have
>> you found that you can build reliable and accurate shifters in
>> max/msp?  i love the fidelity of eventide products but i have to get
>> rid of this eclipse and i'll move to either a bigger eventide or go
>> with something like max/msp.  fwiw, i also use native instruments and
>> i should really look and see what reaktor can do in regards to
>> this...
>
> There is no such thing as an ideal pitch shifter for all purposes. In my
> own music, I am much more interested in structures (time and pitch) than
> in a specific "accuracy" of the shifts. There are basically 3 to 5
> different approaches technically which you could use in Max/MSP. The
> most common one is granular pitch shifting, which is playing short
> grains of the original sound in varying speeds. This is the typical
> sound of "cheap" shifters. I actually use that sort of shifting mostly,
> for two reasons. First I do like the artifacts and it can be set in a
> way to be rhythmically more or less precise. In Max you always have
> access to the small details of such a process (grain size and envelopes
> come to mind).
> Then there is a fft based method coming with Max with less control of
> the detail, as it is within an external, but with very good results for
> sustained sounds (ideal for cello or sustained guitar...).
> And there are two different commercial externals, one is the same
> pitchshifting/timestretching method used in Ableton Live and most of the
> other sequencers and tools. That external will cost the enormous amount
> of 20$ I think. The same programmers also made their own algorithm,
> which might be a better sounding granular approach, but that is
> speculation and solely out of my listening imagination.
> And last but not least, I always wanted to create a sort of very long
> grain looper, which instead of using grains in the range of 50 ms, use
> complete phrases, and break them up at silent parts. This would of
> course scramble the time structure significantly, but would be as clean
> as the good old varispeed...
> This last idea is still on my to do list though...
>
> But maybe you get an idea why Max/MSP is so intriguing as a tool for me,
> there is no technical limitation for your imagination. If you don't like
> the way it sounds, you analyze why and whats missing, and then start to
> invent those missing parts. You are completely free to define your user
> interface yourself and fit it to your personal needs and ways of 
>playing...
>
> Cooking yourself is always better and more fun than heating up
> pre-cooked dishes, even better than going to a gourmet restaurant (like
> your eclipse). Though from time to time its also nice to know that
> terrible expensive cook, he will widen your perspective as well...;-)
>
> This is a long reply now, maybe you want to share it to the list as
> well?...;-)
>
> All the best,
>
> Stefan
>
> --
> Les Ondes Mémorielles--------x--
> -_____-----------|-----------|--
> -(_|_ ----|\-----|-----()-------
> - _|_)----|-----()--------------
> ---------()----------TJ Shredder
> http://tjshredder.wordpress.com/
>
>