Semitone shift for chromatic programs

Anyone (stereroid?) really spend time creating melodic programs?
My real question is this: What's a decent range of semitones in the mpc1k for synths? I use some automated programs to sample sounds from my DMPro, and looking to do the same for some synths.
But what's a reasonable spread for a single sample? ie- good trade-off is using 5 samples across 2 octaves? Is the mpc better at shifting up than shifting down?

I'm playing with this right now, and finding endless combinations, so just wondering if anyone's "been there, done that" already and has some recommendations.

It's a sound by sound situation. Some synth sounds like a pure sine wave or square wave would require only one period of the waveform for all notes.

You'll need to use more than one sample for more complex sounds because when you shift the tuning of a sample with the MPC 1000, you will not only shift the pitch but also the formant and length (time) of the sample.

Examples of incorrect formant shifts can be heard by sampling a voice and tuning it to higher and lower tunings. Tuning the voice up will give a "chipmunk" sound, while tuning the voice down will give a "giant" sound. The time change of a sample can also drastically change what sounds natural. If you're sampling something with reverb or delay and tune the sample higher, the reverb and delay will be shorter and won't fit with other samples running at the original tuning.


Sooo.... it depends on what kind of sound you're sampling. Trial and error might be the best way to go

Thanks stephen, I understand all the theory, and (unfortunately) know the style of the MPC pitch shift algorithmn, and how it does nothing for formant correction (and no single pitch shift method is good for everything either).

But, in the midst of tedious labor, I thought I'd poke & see what people's opinions are. I usually keep my Nord close to the MPC, but wanted to try some other more portable otions--like resampling everything in to the MPC and see how that plays. As usual, nothing is the perfect solution, but somethings are close enough for government work.

the 1000 has pitch shift?

No - it can only transpose.

i think a good standard is usually to have 3 or 4 semi's inbetween samples, but I usually start with one and then work my way up until it sounds good.

I agree totally with what Stephen sez - in general, think about the harmonic complexity of the sound, and sounds with Formants really can't stand being shifted very much.

A few words on Formants for everyone else:
- If you sing at different pitches, only part of your voice's spectrum is being raised or lowered. Part of the spectrum doesn't move (much) naturally, that's the Formant, it's central to the character of the voice.
- Normal transposing or shifting shifts everything, which is why you can end up sounding like Mickey Mouse or Satan if you shift indiscriminately. I have a DirectX plugin on the PC called "RBC Voice Tweaker", which can do the opposite - shift the formant without changing the pitch, e.g. I made Suzanne Vega's "Tom's Diner" sound like it was sung by Henry Rollins.

Hey Stereoroid,
That RBC voice tweaker sounds great.
I have a question, why not a PC application that you can toss a sound in, and it can automatically batch out 63 new samples, higher and lower pitched based on your settings, all time stretched, formant adjusted, and pitch shifted to sound nice? Then if it can also make an MPC1000 program file so you can load it right in, we'd be all set.

Dahkter wrote:Hey Stereoroid,
That RBC voice tweaker sounds great.
I have a question, why not a PC application that you can toss a sound in, and it can automatically batch out 63 new samples, higher and lower pitched based on your settings, all time stretched, formant adjusted, and pitch shifted to sound nice? Then if it can also make an MPC1000 program file so you can load it right in, we'd be all set.

Now you got me thinking and Googling...
I'm confident there are Linux command-line tools for this, and some may run under Mac OS/X or Cygwin*. As for a nice Windows or Mac GUI version, I dunno... let me see.

You could surely getaway with 4 samples per octave, each with -1, 0, +1 shifts - you don't need every single semitone, you'd run out of memory too soon!

* http://www.cygwin.com/ = a Unix shell for Windows, with apps, compilers, the lot.

Java or Perl would be a much better choice for something like this: then you could run it on anything (though I use Cygwin too, and it works for 95% of everything...except pipes & fifos!)

I'm not much of a programmer, but have lots of source & theory on this type of stuff...Is this a side project? I'd kill to automate my sampling & program generation, mostly because I like my studio & all my gear, but hate to bring the studio with me when I play out...I'd much rather grab the sounds with the MPC and just take that into the world.

Please post if you're a programmer, have interest in this, or can otherwise contribute to a project like this. I'll setup something on sourceforge if it's a go.

You'll probably want to start out with a definite problem statement and requirement specification before you begin a project..

Also, you might want to check out SoX:
"SoX is a command line utility that can convert various formats of computer audio files in to other formats. It can also apply various effects to these sound files during the conversion. As an added bonus, SoX can play and record audio files on several unix style platforms."

Some of its many features include:

- Change the pitch of a sound file without effecting its speed
- Stretch/shorten the duration of a sound file
- Change the speed of samples being played
- Trim audio data from beginning and end of file
- Change sampling rates using several different algorithms
- Reverse sound samples

http://sox.sourceforge.net/

As a follow up, I just installed SoX on my PowerBook (Mac OS X 10.3) just fine. I disabled Ogg Vorbis support though. Download the .tar.gz file from the web and put it in your home folder. Then start up Terminal and type:

tar zxf sox-12.17.4.tar.gz
cd sox-12.17.4
./configure --disable-ogg-vorbis
make
make install

and you should be good to go.

ooOOoo This is pretty damn cool. Good idea stereoroid - I've known about SoX for years, but have never bothered doing anything with it. I just made a little script to pitch a sample by a certain amount a specified number of times:

sh pitchem.sh brass.wav 100 11

Takes brass.wav and changes the pitch 100 cents higher for each 11 new samples. In the end I have a folder with a full octave of samples - a sample per semitone

You can download the script here:

http://ieee.usask.ca/members/snorum/misc/pitchem.txt

Sweet link stephen...why disable Ogg?

Looks like a good start...All I'd need is a program to (automatically) trigger my external MIDI device to get multiple samples, capture the audio, Trim/Crop, Remove noise, Noramilize, and build the .PGM file for the MPC. Most of these functions are in someway open sourced, so it can't be terribly difficult to put together.

Like I said, I'd like to just strip a bunch of samples out of my external devices (or even software progs), and munge them for the MPC to use...but do it automatically.

Ogg Vorbis gave me a compile error, so I just disabled it instead of spending time looking at how to fix it . It is probably because I don't have the library installed.

Ok, so what you want is this:
- Program that records audio
- Splits the audio file into separate segments for each sound
- Maps sounds to MPC 1000 program file chromatically
- Tunes samples required amount in the PGM file

?