Quote Originally Posted by The Man View Post
Well, the problem is, how do you <I>measure</I> what the difference in sound is between different types of fingers playing the piano, and how do you generate an algorithm that will accurately model the difference between different hands?
Not disagreeing that there aren't a lot of hurdles to overcome before it's feasible.

That synthesizers will eventually be able to come up with any number of subtly different sounds I don't doubt, but will they accurately model what would happen in the real world? It's much the same issue, I think, with models of the environment:* There are more variables at hand than we can possibly hope to incorporate into any humanly generated model. Until we actually come up with a way to write a computer program to model every aspect the physics of playing a piano, it's unlikely that anyone will be able to come up with a synthesizer recreation that feels right.
I'm not convinced the question is necessarily one of needing to model it accurately to do it convincingly. Yes there are a lot of variables at play, but is it necessary to model them all to fool a human, even an expert? Maybe not. It's a tough question to answer but we'll probably see people take it in both directions in time.

And it's certainly conceivable that such a thing <i>could</i> be done eventually, but it begs an important question: Where's the profit in it? I guess it's conceivable that a person might do it entirely for the intellectual rewards, but it seems like it would be an extremely costly project and unless there were some guaranteed way to at least recoup the amount of money required to research the question, it seems unlikely anyone would be willing to pursue it.
You may be somewhat surprised there. With regards to physical modeling at least it will always have value in a production sense in that it can be used and manipulated to provide a great deal of control over a sound. The latest version of the software Reason for instance includes a software instrument for creating drum sounds using physical modelling. I haven't used it so I have no idea how accurate it actually is, but just the same it has value for those who like tweaking the sounds they use to no end. And given that there's no shortage of virtual instruments of all kinds out there, someone will do it if they think they can improve on what's come before, and they probably will find an audience for it. That kind of iteration and experimentation in the long run may lead to something that accurately produces the results if not the variables that cause them in a real instrument.

Now creating more robust sensors and translating that input into the model to produce a believable sound is another matter, but there certainly is a market for it. If someone created an electronic drumset that replicated the sound and feel of performing on an acoustic kit to the degree that I couldn't tell the difference, I may be prepared to ditch acoustics. At the very least, I would never lug an acoustic kit anywhere ever again unless I was moving. The value of having a realistic sounding and performing instrument that's a fraction of the weight and takes up significantly less
space is something I don't think should be discounted where some instruments are concerned, and there's enough competition in the market for these that there is certainly a profit motive to drive further R&D. Whether that R&D takes them in that direction though is something we can't really predict, but the consumer demand for more realistic electronic instruments is something I think is there.

Of course, like anything, this is all speculation. My main point was simply that saying it will never happen strikes me as more than a little short sighted.