Behavior Modeling Explained -

Forum for JUPITER-80
vladuca
Posts: 28
Joined: 08:21, 7 April 2011

Behavior Modeling Explained -

Post by vladuca »

I'm working on a document explaining Behavior Modeling, and thought to share this rough draft with the Clan. This is not the official explanation as I still need to add a bit more about SuperNATURAL piano and EP, but you'll get the idea of what we're up to, minus the napkin drawing!

++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

What happens when a keyboardist is trying to reproduce sounds other than a traditional keyboard instrument on the synthesizer? This is where synthesizers today really fall short because the typical sound engine inside cannot truly “interpret” an expressive keyboard performance done on keys into an authentic guitar, trumpet, or violin performance – a guitarist plucks or strums strings, a trumpet player uses his breath, lips and trumpet valves, and a violinist plucks or uses a bow on strings. All the keyboardist can do is trigger a static digital sonic picture of sound he is trying to recreate, and each time a key is pressed, the same sonic picture is repeated, but at different pitches as played on the keyboard.

The Jupiter-80 solves this problem by using Roland’s newly developed Behavior Modeling technology. It takes care of the “interpretation” by constantly analyzing the keyboardist’s normal, natural keyboard performance, and instructs the ultra-realistic SuperNATURAL™ sound engines inside the Jupiter-80 to “play” and constantly “articulate” the reproduced sound just as the “real” performer would. These articulations are based on the keyboardist’s timing and interval between notes, the strength at which the various keys are struck, or if the keyboardist is pressing a pedal to sustain notes. These actions, among others available to the keyboardist, are translated by the behavior model for the selected sound being reproduced into an authentic plucked or strummed performance in the case of a guitar, aggressive or smooth bowing in the case of a violin, and the sharp or smooth pitch changes created by the valves on a trumpet and the pressure of the trumpet player’s breath. These are but a few of the articulations possible with Behavior Modeling.

Using the analogy of an orchestra - the keyboardist is the conductor, his keyboard playing is “reading the score and waving the baton”, and the Behavior Modeling technology is the player in the orchestra pit, and the SuperNATURAL™ sound engine is the actual instrument. Just as a conductor receives feedback from the orchestra depending on his baton articulations, the Jupiter-80 also gives feedback to the keyboardist. This feedback loop and exchange between keyboard player and instrument is akin to a real acoustic instrumentalist’s experience, and allows the magic of becoming “one with the instrument” for the performer, and creates complete suspension of disbelief for the audience as they look on stage for the “real” instrument.
Jan_nl
Posts: 203
Joined: 10:10, 2 May 2008

Re: Behavior Modeling Explained -

Post by Jan_nl »

So, from what I read behaviour modeling is the equivalent of what for example VSL allows us to do with their player, i.e., triggering the right samples depending on the kind of input, is that correct?
Is there more to it in the sense that the sound of the basic samples is being manipulated, or is all this purely sample based?
Undercoverman
Posts: 69
Joined: 01:06, 17 August 2008

Re: Behavior Modeling Explained -

Post by Undercoverman »

Thanks Vince.

Those of us who've worked in the world of software instruments have had access to similar technologies for a while now, and let me tell you - it makes a huge difference. We're getting to the point where what makes a difference is not the size of a sample library (as you're probably aware, sample libraries for particularly expressive instruments run up to hundreds of GB of content), but how well it has been programmed to include things such as legato detection, chord detection, strum direction and string selection. A software/hardware synth instrument emulation that best responds to the intentions of the player is a wonderful and, more importantly, a rare thing. Hearing a well executed slide between two notes on a string instrument emulation played via the keyboard is genuinely surprising when it first happens without benefit of a bend wheel, or seemingly in the right spot.

I do actually think Roland has got this part of developing the sounds in their instruments exactly right, because they've recognised that great programming is always going to be more important than the sample size. This is an important step forwards (for Roland).
Undercoverman
Posts: 69
Joined: 01:06, 17 August 2008

Re: Behavior Modeling Explained -

Post by Undercoverman »

Jan_nl, from what Vince is saying it seems like the behaviour modelling is similar to that which can be achieved with, for example, heavy scripting in Kontakt. I think there's two types of modelling in the Jupiter 80 - actual instrument modelling with algorithms calculating the sound, and behaviour modelling based on the input from the player. Think of an instument like Musiclab's Realguitar (behaviour modelling, sample based) versus Applied Acoustic's Lounge Lizard (physical modelling, no samples). I think the Jupiter 80 contains examples of both.
vladuca
Posts: 28
Joined: 08:21, 7 April 2011

Re: Behavior Modeling Explained -

Post by vladuca »

I think there's two types of modelling in the Fantom G
I think you mean the JUPITER-80 :-)

...and the cool thing is no scripting! Just play!
Undercoverman
Posts: 69
Joined: 01:06, 17 August 2008

Re: Behavior Modeling Explained -

Post by Undercoverman »

Haha, did I really do that? My god. Force of habit, sorry!
vladuca
Posts: 28
Joined: 08:21, 7 April 2011

Re: Behavior Modeling Explained -

Post by vladuca »

Thanks for the edit! We wouldn't want to start a new rumor now... LOL
User avatar
V-CeeOh
Posts: 3956
Joined: 18:13, 28 September 2004
Location: Portugal

Re: Behavior Modeling Explained -

Post by V-CeeOh »

We wouldn't want to start a new rumor now...
No, no, no, please do ;-)
en76
Posts: 664
Joined: 06:57, 3 June 2008
Location: honolulu' Hi

Re: Behavior Modeling Explained -

Post by en76 »

I guess it would be accurate to say This JP-80 simply inspires me in a way that we are one....:D so good to know and good to be on the right side of moving forward. Thanks 4 the explanation. This exactly why I purchase the FG to begin with. Is how realistically a keyboard can play such tone of a certain instrument to almost perfection/identical and having people looking for the real thing on stage.

The idea of perplexing the audience to look for such instrument that physically absent and having your listeners" guessing which sax player you hire" or "imagining a full band present with" you but yet "none" really satisfying to me.
Well, on the other hand, I don't need to hire a solo player when your doing your composing.
Beside, who wants to be left out on the fun/thrill of jamming away on such tone of instrument without the actual instrument? For sure not me...LOL!
Well, most of all is that Roland is on the right path which is exactly what I am looking for and willing to spent on. As long as they keep it supernatural/behavior modeling, of the best acoustic instruments out there, I am on board.
Now this kind of Vision what keeps a company a float and successful-Daring and innovative.
This means that roland basically have the whole world to work with. Acoustic sound/instrument is just basically all over the world.
Endless work for roland with this technology. almost never a two acoustic instrument have the same tone.

This is a gold mine for Roland. Hope you scout out the whole world for that unique sound/instrument out there.
vxfan
Posts: 42
Joined: 00:49, 18 December 2009
Location: USA

Re: Behavior Modeling Explained -

Post by vxfan »

With Behavior modeling and Supernatural, I wonder how many samples are in a specific SuperNATURAL sound that the behavior modeling technology controls?
Jan_nl
Posts: 203
Joined: 10:10, 2 May 2008

Re: Behavior Modeling Explained -

Post by Jan_nl »

Undercoverman wrote:Jan_nl, from what Vince is saying it seems like the behaviour modelling is similar to that which can be achieved with, for example, heavy scripting in Kontakt. I think there's two types of modelling in the Jupiter 80 - actual instrument modelling with algorithms calculating the sound, and behaviour modelling based on the input from the player. Think of an instument like Musiclab's Realguitar (behaviour modelling, sample based) versus Applied Acoustic's Lounge Lizard (physical modelling, no samples). I think the Jupiter 80 contains examples of both.
Thanks for your reply, Undercoverman.

I managed to find some more info which made it clear how Supernatural combines a sample with modeling techniques at Roland's website.
How exactly this works is not clear, but I suppose it runs parallel to what Samplemodeling does (i.e. sampling with modeling techniques such as harmonic alignment and convolution techniques). What I do know is that this marriage between sampling and modeling is capable of producing extraordinary results:)
kenchan
Posts: 1876
Joined: 23:46, 22 December 2008

Re: Behavior Modeling Explained -

Post by kenchan »

so is this like the VL synthesis yamaha did like 15-20yrs ago?
User avatar
Artemiy
Site Admin
Posts: 19754
Joined: 13:00, 17 April 2003
Location: Ukraine
Contact:

Re: Behavior Modeling Explained -

Post by Artemiy »

kenchan wrote:so is this like the VL synthesis yamaha did like 15-20yrs ago?
VL is physical modeling. Like in Apple's Sculpture synth, or Pianoteq virtual piano.

Here we're talking about modeling of note transitions, pitch bends, various other playing nuances, i.e. not a math model of the sound source itself, but how it's played. The source itself is recreated by SuperNatural which is infinite morphing between multiple sampled layers.
kenchan
Posts: 1876
Joined: 23:46, 22 December 2008

Re: Behavior Modeling Explained -

Post by kenchan »

so we can all sound like we know how to play keyboards now? :D
Leh173
Posts: 513
Joined: 06:08, 8 August 2010
Location: Sydney, Australia

Re: Behavior Modeling Explained -

Post by Leh173 »

The specs on Roland's site lists this: Harmony Intelligence, 17 types. If a keyboard can have 17 types of Harmony Intelligence maybe we could programme some into people so the can be be more harmonious too. LOL. :-P
Post Reply