posted 01 March 2012 07:43            

But for a kyma users nothing to special there. The featureset, userinterface and performance features are very well thought out and designed (thanks to Errorsmith). It makes fun using it. anyway it uses lots of CPU resources.

NI started to "protect" the reaktor ensembles you buy now ... and you are not allowed to look inside and make changes. A "commercial decision" which is the END for my support of the Reaktor community.
Copy protection YES but restricting for the buyer NO way!

my 2 cents..

all the best
Chris

posted 13 March 2012 15:40            

Now also forget Capytalk as this is good for control time (1khz refresh) stuff but here what you want is sample rate type control.

So what you need is to knit modules together instead.
First thing is to start with your spectral analysis (live or pre analyzed) feeding into your oscillator bank. Then separate the left and right signals to two different paths and re join them. Putting the Channel Jion prototype between the two modules will do most of the work for you, but you must clear out the delay module that you get for free in the channel join prototype.

Then put a product (multiply) module in the left only leg of this split path. Then with a ramp made from an oscillator module with ramp as the waveform with a repeat rate exactly matching the frame rate in samples of the analysis, feed this into an InputOutputCharictaristic module which in turn feeds into the same product module mentioned above. Now by putting hot values into the inputOutputCarectaristic module you can control the levels of the partials as a form of filter.

If you make the InputOutputChar module produce output value 1 regardless of input ramp value you will get the original audio signal unaffected as you are multiplying all the partials levels by one. If you instead make it 1 to 1 so that the ramp goes straight through the InputOutputChar, then the overall effect will be a hi pass filter as the highest partials will be multiplied by one, but the lowest partials will be multyplied by zero. As you make different shapes by varying the values in the InputOutputChar module you can make very complex shaped filter shapes. These are controlling the levels of the partials.

I won't go into detail here as to the values you should use, but if you think about what is happening you should be able to work it out.

Hope this makes sense.

If you have seen the video of my talk at the KISS 2011 the wire between, well the wire I'm talking about is the wire you would have put the product module in. This video may help a bit.

Pete

posted 17 March 2012 11:58            

508partialmodificationV2.kym

I hope you don't mind but I made some changes.

I didn't have your SOS file so I used one from the Kyma library instead.

First because the ramp was a half ramp, the first 2 controls were doing nothing so I've changed the values in the I/OChar module.
I changed it to 512 partials just because the scope is a bit more stable (not much mind).

I've added the ability to move the points between the bands and I've set these as log controls. This is because nearly all the noticeable change in sound is are the lower harmonics.
When tweaking these controls you will need to make sure that they stay as incremental values from the left to the right or some of the level sliders will have no effect.

I've made the faders positive only, as negative values just turn the partials on again.

Hope it does what you wanted?

posted 03 April 2012 10:32            

I have posted it up at the Share on the Twiki ..

posted 06 April 2012 12:13            

http://www.symbolicsound.com/cgi-bin/bin/view/Share/Sounds#Spectral_processing

posted 07 April 2012 11:06            

In this case the right hand side goes straight through and is left untouched.

Note that this was a simple example of adjusting the levels of the partials but you can push the level envelope (spectral envelope) between partials and cause formant shifting.

Playing with the right hand signal will change the pitches of the partials. There are a few considerations here. If the spectrum is harmonic, the pitches will have a strong relation to the pitch of the signal, but if they are non harmonic, the partials keep within there own bands and the pitch movement is very small. The pitch of the input signal in the latter case is more related to the levels (left side) as you will see increases of level in the band that relates to the pitch of the content.

Also live spectrums tend to have the (right leg) pitch info represented in lin spacing, where as spectrum in ram (pre processed) tend to be in log spacing.

There is a lot more detail in the video of my talk at KISS 2011, but although it is impossible to see the scope displays on the screen, you can request the accompanying kyma sounds and play them along side the video.

Request the details here.
http://www.symbolicsound.com/cgi-bin/forumdisplay.cgi?action=displayprivate&n umber=6&topic=000472

and the video can be found here
http://kiss2011.symbolicsound.com/videos-photos/

look for "15:00 The Piece of Wire Between, Pete Johnston"

Please let me know if this makes sense.

Pete

posted 07 April 2012 16:22            

Email with link should be with you now.

Pete

posted 08 April 2012 02:24            

The live spectral smearing sound has a LiveSpectralAnalysis module with the LowestAnalyzedFreq parameter set to 1F (44 hz) and produces a 512 partial spectrum.

If you change that parameter and leave the delay (labeled "smearing" in the sound) set to 512 samples you get an extreme smear effect because all of the other LowestAnalyzedFreq parameters produce a number of partials divisible by 512.

posted 12 April 2012 08:18            

Just one additional question: suppose that I want to shift all the partials in the sound by a fixed amount of 100Hz (just to use as an example). In this case, what value should I add in the right hand signal? Or put in another words, how do I translate the values in the right hand signal (that I suppose are in the range [-1; +1]) to a frequency in Hz?

Best regards,

Jose

posted 12 April 2012 16:28            

Well there's two answers as it's depends on Log or Lin frequencies.

Don't forget that normally spectrum in ram gives frequencies in Log and live spectral analysis give frequencies in Lin.

I hope SSC will correct me if I'm wrong, but from memory in Lin the range is 0 to +1 where 0 is zero hz and +1 is half the sample rate (22.5 khz when 44.1 hz sample rate). So to increase by 100 hz you add (100/22500) to the right leg.

Log is quite different and not so easy. the range is 0 to +1 where +1 is half sample rate and the range is 9 octaves. So 0 is 22500/(2^9). Simply adding or multiplying will not give you a fixed increase in hz.

Most of the time you don't want to add a fixed frequency as that will put the harmonics out of tune with each other. If you wanted to change the pitch and keep it sounding tuneful (not like a ring modulator) then you would use the following.

For Lin multiply by 2 for an octave or multiply by ( (1/12) twoExp ) for a semitone or ( (2/12) twoExp ) for a tone.

For Log add 1/9 for an octave (double the pitch). For a semitone add 1/(9*12) or 2/(9*12) for a tone.

I hope this makes sense and I hope I've got it right.

posted 20 April 2012 07:29            

Thanks to all involved!