EvilDragon wrote:Kevin Nolan wrote:Supernatural is a family of synthesis approaches; all using computer modelling. It's not scripting.
Actually, it's based on samples still. It's not pure modeling.
Kevin Nolan wrote:For Supernatural Acoustic Tones, it's a combination of samples and "behaviour modelling" - models on how instruments sound when performed in given ways. It goes beyond the likes of Kontakt scripting or VSL Instruments because it does not use velocity switching of samples. Instead, the "Supernatural Acoustic Engine" responds to playing and MIDI, and interprets that as a behavioural model applied to the samples. What actually goes on within the model is not clear, but it must involve a number of techniques ranging from formant synthesis to realtime convolution of instrument behaviour to performance types."
This is EXACTLY what Kontakt scripting can also do! So, Supernatural doesn't really go beyond the likes of Kontakt in this regard. Convolution or formant synthesis have absolutely nothing to do with it. And yes, they ARE using velocity crossfading.
And I do Kontakt scripting for living, so I know a thing or two about it
Here's a showcase that it's all possible with a single patch and some modwheel right in Kontakt:
http://spitfire-webassets.s3.amazonaws. ... emo_AB.mp3
Spitfire Artisan Cello and Violin. To me this completely kills any Roland's attempts, really. It just sounds so much better.
Well, the sound source is samples - but - the Behavioural model is that - a computer model. It's "Behavioural Modelling". It's a 'model' is in the computer sense - absolutely it's not scripting - rather the behavioural characteristics of performance are computer modelled in real time - and encoded in a chip. And I'm saying in this instance there is very likely both formant synthesis and convolution going on having played with it and from my experience (physics degree) - Roland are not saying that but there has to be such processes going on as part of it (including many other DSP techniques of course)). There's nothing surprising in this - I'm just saying it to highlight one or two of the differences between this and "scripting" which is a scripted and interpreted (at the OS / Application level) list of instructions for switching in and our samples.
For example, they manipulate the actual sampled sound in real time, using formants, like Variphrase on the V-Synth GT - to achieve the actual sound from the original sample as it should sound when performed a particular way. So they manipulate the samples in realtime - not switch in new ones (in some instances). You can actually hear it in some instances. There's (extremely likely) convolution going on too because for certain sounds they merge the sound with the soundboard of the instrument - requiring realtime convolution (where convolution means multiplying each sample point of one sound source with every other sample of the other sound source - in real time - at ( 44.1/ 48 ) kHz).
Scripting is a totally different phenomenon. What's going on there is that a number of commands are _interpreted_ in real time at the application/ OS level to instruct Kontakt to switch samples -very much like UNIX Shell scripting. It's literally millions of times slower. I have both the JP80 and Kontakt with scripted libraries, and I can assure you, they are utterly different. You can achieve some reasonably respectable results with Kontack - but I assure you - the JP80 is in a totally different league in this regard. At times its extraordinary.
You really do need to spend some time with the JP80 to realise the sophistication - nobody's saying its Instrument Modelling - rather - it's Behavioural Modelling - and it's extraordinary.
THE central point being however - the sound of the instrument is always dynamic and always different - depending on how you play it - because its computed in realtime. The nuances are extraordinary. For example, you can play two notes at precisely the same velocity on a solo violin, and for one, play staccato to the next note and for the other legato the the next note. The resulting over all sound is utterly different in each instance. For the staccato scenario not only do you hear both notes fully articulated from scratch while for the legato notes you don't hear the attack of the second note, but the combined character of the both sets of notes (staccato or legato) is utterly different - and sounds just right! There's a huge amount going on - because it's being computed in real time. It's a realtime model, computationally. In this regard it is not _totally_ different to the VL1 because although the VL1 does not do behavioural modelling, the instrument models are _so_ complete they do include some behavioural characteristics during performance so that if you play from note to note for solo instruments in various performance modes, the instrument model behaves accordingly - the classic example being the creation of "edge harmonic tones" between two flute notes if you don't get the timing right between lifting one key and depressing the other.
I'm afraid, as goos as scripting can be, it comes no where close. The JP80 models are, as said, totally dynamic, computed always in realtime, and the sound is different in every different playing moment - Roland have done an unbelievable job and is surely the result of years of research. Such is the backlash against the Jupiter name that this technology has been beaten out of existence - but there are those of us who are exploiting it. I see it already as " a best kept secret" - and I assure you - just like the VL1 and EX5R - I fully expect the JP80/50 and Integra to establish a delayed new level of respect in perhaps 5 - 10 years from now as people begin to realise just how capable these instruments are.
I said it feels like cheating because there are times when you're playing up the keyboard with some instruments you hear this "performance" emerge that you realise you're not totally responsible for - it's great for media work - but where your "soul" ends and the JP80 takes over is often not too clear
@Bachus - while I broadly accept what you are saying - actually - not all synthesis is as you describe. While I personally largely subscribe to your idea - many don't. The origin of the Fairlight, Emulator and M1 were largely motivated by a desire to reproduce acoustic instruments, not create new ones.
And in fairness to Roland - they have been consistent - since their origin - with an obsession to recreate real instruments synthetically. Their founder was literally obsessed with this challenge. Hence the long lineage from the VP330 to the V-Synth GT to recreate the human voice both chorally and through vocoder.
And while Roland were attacked for using the Jupiter name for the Jupiter 80 - Roland claimed that it was legitimate because the instrument's aims were the same as those of the Jupiter 8 - to recreate real instruments. People largely do not accept Roland's word on that - but I went back and checked - and sure enough - all of the blurb at the time of the Jupiter 8 emphasised it's aims to recreate realistic brass, strings and other acoustic instruments, and NOT as a deice to create new sounds.
Of course the Jupiter 8 is not known for that and has become a legendary analog polysynth - but - Roland's goal was to create instruments that emulated the real thing. You and I may not like it,but Roland HAVE been consistent in their ethos - from the 70's right up to the JP80.
I also accept your idea that in some ways the JP80 is a retrograde step from the V-Synth GT. The reason - the V-Synth GT also has a very similar model to Supernatural Acoustic Tones called Articulation Phrase Synthesis. It too is a behavioural modelling approach (and surely lead to Supernatural Acoustic Tones). And while on the GT they only provide 3-4 models (Violin, Sax, Erhu and one or two others) - you can attach those behavioural models to other wave forms - including the VA waves such as Sawtooth. So you can play a Sawtooth lead line with the expressivity of a Violin, for example ! You cannot do this on the JP80 - each behavioural model is firmly attached to it's sampled instrument and there's no way round it!
