KVR Audio

Gregjazz · Post by **Gregjazz** » Thu Aug 13, 2009 9:46 pm

As some of you know, I was involved in some physical modeling projects a long time ago, working on physically modeled basses and guitars. I used Synthedit, which while being powerful, was also very CPU hungry and impractical for big physically modeled projects. I ended up learning some C++ so I could make my own custom modules like Guido Scognamiglio from GSi does, but my projects still didn't get very far.

I've been involved with creating sample libraries for the past decade, and now run my own business, Orange Tree Samples. My experience with physical modeling showed me the short-comings of traditional sampling, so using Kontakt's powerful scripting engine I am now able to apply my knowledge of physical modeling to add realism and playability to my sample libraries. While the samples themselves are recorded from real instruments, the context in which they are played are modeled after the behavior of the instrument as well as its player.

Yesterday I had the crazy idea to try pure physical modeling in Kontakt, and after a quick test, I came up with this: http://www.orangetreesamples.gregjazz.c ... taktPM.mp3

It's really rough, modeling a single wave on a string, but it proves that Kontakt IS capable of physical modeling. The CPU usage isn't bad, either!

The next step for me would be to make the wave bi-directional instead of uni-directional (making it possible to model the pick position on the string) as well as adding pickup modeling.

Krakatau · Post by **Krakatau** » Thu Aug 13, 2009 9:54 pm

Gregjazz wrote:It's really rough, modeling a single wave on a string, but it proves that Kontakt IS capable of physical modeling.

really astonishing....

...but how is it possible, Kontakt scripting capabilities ?

Gregjazz · Post by **Gregjazz** » Thu Aug 13, 2009 11:18 pm

Krakatau wrote: really astonishing....

...but how is it possible, Kontakt scripting capabilities ?

Thanks! Yes, it's all Kontakt scripting!

marce · Post by **marce** » Fri Aug 14, 2009 1:26 am

so the sound source of this mp3 is a math formula or so and not a sampled wave file?

mojamusic · Post by **mojamusic** » Fri Aug 14, 2009 2:29 am

Greg,

you are a mad scientist

Keep up the great work.

So... I can't wait to see where you take this!

Myron

Gregjazz · Post by **Gregjazz** » Fri Aug 14, 2009 5:18 am

marce wrote:so the sound source of this mp3 is a math formula or so and not a sampled wave file?

Yup! Like I said, it's a really quick, rough test, so the audio is not representative of the final product, but it demonstrates that pure physical modeling in Kontakt is indeed possible and realistic CPU-wise!

Ben H · Post by **Ben H** » Sat Aug 15, 2009 2:24 pm

Wow, that's very impressive.

Even in it's most basic stages it sounds quite promising.

Jake Jackson · Post by **Jake Jackson** » Sun Aug 16, 2009 11:49 am

Looking forward to what you develop. What kind of hit is the cpu getting?

Are other people doing physical modelling in Kontakt?--I think "The Trumpet" was done in Kontakt. Or did they use samples and modelling together?

Krakatau · Post by **Krakatau** » Sun Aug 16, 2009 12:18 pm

Jake Jackson wrote:Looking forward to what you develop. What kind of hit is the cpu getting?

Are other people doing physical modelling in Kontakt?--I think "The Trumpet" was done in Kontakt. Or did they use samples and modelling together?

+1

Interesting point, i'm also curious to know if they are inner limitations of the technique within Kontakt (... or not !)

freddemillio · Post by **freddemillio** » Sun Aug 16, 2009 1:02 pm

That is quite promising sounding and I am looking forward to see where this leads you (and then where it leads us

).

Thanks for sharing a glimpse of your work in progress.

Gregjazz · Post by **Gregjazz** » Sun Aug 16, 2009 4:31 pm

Jake Jackson wrote:Looking forward to what you develop. What kind of hit is the cpu getting?

In the example in the thread, I'm getting 1-2% CPU usage, which is really good considering the processing.

Jake Jackson wrote: Are other people doing physical modelling in Kontakt?--I think "The Trumpet" was done in Kontakt. Or did they use samples and modelling together?

The Trumpet is sampled. However, they use a trick to make the samples more flexible, by phase-aligning all the samples and then crossfading between them, they can get all the dynamic ranges, various different attacks, etc.

D.Wako · Post by **D.Wako** » Sun Aug 16, 2009 6:44 pm

...it demonstrates that pure physical modeling in Kontakt is indeed possible and realistic CPU-wise!

Very good news!

deastman · Post by **deastman** » Sun Aug 16, 2009 6:56 pm

Very intriguing! I'll look forward to seeing how this develops.

Gregjazz · Post by **Gregjazz** » Mon Aug 17, 2009 5:43 am

Wow, thanks for all the support! Seriously, this was a really quick, rough test. I'm going to have to work more on this and see what I can come up with.

I made the same model in SynthEdit, with movable pickups and picking position, and I got it to the point of realism where the sustained notes sounded really close to the ones on my guitar. I even compared them in a frequency analysis, and they were practically the same, although the real guitar had more background noise and buzz.

I'll keep you updated with more audio clips and info!

marce · Post by **marce** » Mon Aug 17, 2009 2:21 pm

Gregjazz wrote:
I made the same model in SynthEdit...

and the CPU use was...? Just wondering, since you said that in Kontakt it was very low.

Physical Modeling in Kontakt! (UPDATED audio example)