At FITC Amsterdam I demoed my first and fairly simple implementation of a Karplus-Strong implementation. Together with KP, we were able to enhance it to sound more like a real guitar. I think, the result is already quite impressive, considering spending just a few hours. At some point I cannot hear a difference to a real guitar. This is even more impressive when you think of how it is accomplished.
A simple low-pass filtered feedback in a short delay line with the length of one period of the target frequency and some white noise do the first impression of a plugged string in common. For a guitar you need more tweaking, randomizing parameters and adding the guitar body resonance to get it more realistic. But the basic is done with an algorithm, discovered 1978!
I really like the elegance of the circuit. You basically bring in chaos (white noise contains all frequencies) and the system is going to stable itself and produces a nice waveform, perfectly suggesting a plugged string.
Naturally we are already planning an AudioTool plugin for synthesizing different guitars. But I am worried about the footprint of the algorithm. I stumbled over this comment which pointed out that the Karplus-Strong algorithm is patented. That is tough. At the time being this algorithm was a great accomplishment! But now where even Flash is able to playback code-generated audio data, I would think that someone may stumble over this method even by accident. It is better to commit our lawyer with this issue, but as the comment says, it is a really good example how bad code-patents are.
Listen and picture it for yourself. (Check the CPU!)
The implementation has a static set, but gives you some control on how the guitar is played. Nothing really magic. Definitely it would be a perfect addition to the AudioTool.