Where is physical modeling?

[astroman]

you bring it to the point Eliam, it's all in the interface.
I'm using the STS 3000 only with readymade libs and never attempted any recording myself.
Recently I've experienced how much fun that can be - but guess with what a Casio toy keyboard (SK-8, a whooping 4-voice poly).
The interface is so simple stupid that you have results in a second. OK, somewhat limited results, but it gets you straight to have fun with sampling and just try out what comes up in your mind - 9.38 khz is a nice bitcrusher btw.
Just push the 'sampling' button while that cheesy unit is drumming and sequencing along, it will interrupt, let you shout something into the internal Mic and continue immediately after it got the sample (guess it autocuts by envelope). Push 'sample effect', then one of the keys for loop, envelope or reverse - it will crash a cymbal to inform you it's done the job and hits 2x rimshot to indicate a bit more time needed for reversing.

Very interesting ideas in the above mentioned thread, Eliam, still have to read the details.

I see neither sampling or PM as a replacement for 'real' musicians, for that expression and 'flow' that cannot be matched by any synthetic system.
Of course it's possible to model everything. You don't need that much calculation power, but the right params in the formula.
Like a fractal description of graphical items: you could code the image of a complete city on just one A4 size piece of paper, if you only knew the right params...

But what's all the efforts for ? Most 'consumers' do just that, they consume and don't care how it was produced.
Half an hour in a high-fi department of an arbitrary warehouse or in the middle of a home cinema system will take any illusions from you... And imho for that group most of the commercial music is produced.
If your destination system is MP3, anything goes.

cheers, Tom

[eliam]

I think that this little keyboard you mentioned provided my friends and I with LOADS of fun and hours of rolling on the floor because of too much laughing, a few years ago!!! It's funny that you talk about it now! We were doing parodic techno house caricatures, something like: poom-ts-poom-ts-poom-ts-poom-ts- Hands up! You're busted! -poom-ts-poom-ts-poom-ts-poom-ts... and so on... That song was called Technopolis (technopolice)! We so much laughed that we almost had a near-death experience!

About modeling, I think you're right, all it requires are the proper algorythms and then the most complex structures and textures can emerge, but our machines are still too slow to treat the data in real-time, especially if there are co-evolving formulas which interact with each other, like strings resonating with each-other and with the feed-back from the room, etc. But I trust we will see this happen in not so long! And the interface has to follow for us to exploit the potentialities of virtual instruments. About the above-mentioned thread, I'd like to synthesize (resume) these ideas because all that was said is still not completely clear for me. The more I think about it, the more I wonder how we could get as close as possible to the real expressive properties of let's say a violin, with something else than a violin! It might be more satisfying to play violin than any kind of emulation... Ok, there's the point of inventing unknown virtual instruments, but then, why not create them physically and play them!

Anyway, my personal goal is to learn to play as many instruments as I can and be my own orchestra-nothing less! Samplers are useful to get a glimpse of how it could sound, but the real thing is: acoustic instruments and human voice... imho.

[liulai]

Hi there,

Thanks for your contributions. As we all agree, control interface is important, but that doesn't mean PM is limited. It means PM is limitLESS and needs be unleashed under a control interface that can be easily understood by most members of the music community. Academics have a special grasp on this subject, but the end user is definitely the music performer/producer.

Few can acknowledge this unless they've been into the academic side of it, but - believe me - PM is already 99% sense-fouling with many acoustic instruments. Woodwinds, reeds, basses, chromatic percussions can be modeled just perfectly, I don't mean the old Yamaha VL1 but some experimental systems that are running in some Electronic Engineering department in Stanford, Bristol (UK) or Helsinki - obscure to many of you.

The problem is just that - to bring this technology to the public in a way that won't seem a nasty ride into programming - sound programmers have definitely had enough of that with FM.

With PM you don't have 135 attacks, 25 sustains etc. - you have a live instrument which only boundaries are those imposed to the physical model in order to resonate (and perhaps not even those ones, in the most experimental instances).

One solution could be intelligent randomisation. If I play a sample note with the same velocity - chances are my AKAI patch will playback the same sound at the same amplitude. With a PM flute, I want something slightly different every time. How different, only a pro flute player can tell. So what about getting a sound programmer, a pro flute player and a couple of Yamaha engineers sitting down for some time and understanding how to unleash this into a 2, maybe 3-dimensional control space?

Hope I added something to the discussion.

Laters,
Vanni

<font size=-1>[ This Message was edited by: liulai on 2002-05-07 09:31 ]</font>

[eliam]

You sure did, Liulai! Have you had the chance to hear these experimental virtual instruments? This is really, really interesting...

While a randomize function can certainly be useful, I don't think it would be appropriate all the time. It could generate something to start with and refine after. A useful tool would allow us to link the virtual instrument with an envelope controler or something similar. In this way, we could draw the evolution of the note that we want for each note and have the modeler play accordingly. Actually, I believe we will need a multi-parameter window through which we would input the info. Example: the violin: we must be able to say precisely to the Vinst how are these parameters:
-Bow pressure, bow speed, bow direction, angle of attack, location of attack on the string(how close to the bridge), vibrato(width, speed, evolution in time, etc.), etc, etc. And the evolution in time of all these parameters.
Now, the idea would be to do the same thing as let's say the modular synth, where there are multiple parameters to play with as well as many ready to use presets. In that way, everyone is happy, the archigeeks as well as everyone else!
But this is not a physical interface, but the params could be assigned to various external means of transmitting data to the machine. I don't know much about what exists in this domain...

Very interesting discussion!!!

[astroman]

hi Vanni,

in technical terms you're right, we already have the technology, though it still lacks the interface to use it's full potential (maybe just a question of time).
But you completely miss the personal aspect, the challenge for the player, the unity of movement and acoustic experience. Must be like heaven if you worked hard on a piece for a long time and finally succeed.
This is a powerful driving force for many artists. From traditional instruments to the turntables it's somewhat the same, just the performance varies.
That's why my suggestion is to completely give up imitating instruments by PM for that purpose.
To use the sound in the studio is ok, there are numerous applications where it doesn't matter if a tune is playable on a real instrument.
Devices like 'Surface One' (if it ever gets released) by Midiman give an opportunity to develope new playing styles, so why not take the chance ?

Eliam, the fun factor with the Casio was my first impression, too
But it really changed my view of samplers as creative tools. After the jokes of shouting at the thing, I actually SAMPLED some Pads and a few bars from a drummachine.
The toy made me turn my ideas to sound in just a few seconds and I was quite impressed how the onboard capabilities were extended.
I consider myself knowing a lot of techie stuff about samplers, but that doesn't help with inspiration.
It's really different doing it, instead of just talking about, btw a familiar experience...

The pic example reminded me on fractal image representation, where you can define a nearly infinite complex scene by an actually very SIMPLE calculation - you just need the right params, but those are obviously nearly infinitely difficult to find.
It seems to me that the more processing power is available, the more complex is the WAY of calculation, but still the same way.
There's a good chance to have success on the opposite side with simple calculations yielding sophisticated results, as mentioned above.

cheers, Tom


<font size=-1>[ This Message was edited by: astroman on 2002-05-07 17:43 ]</font>