Trying to modulate parameters with velocity

Well, my impression is that many of the parameters are of course already dependent on velocity, so just the sliders can be used.

So my first goal was to try to modulate the Unison detune inversely with velocity. In other words, to make the unisons come more into tune with harder strikes, thus trying to replicate their coupling. 

But I also started thinking along the same lines that you have been thinking. I don't know if the partials in a real piano, or in the spectrum profile, all react linearly to velocity. Obviously, the higher partials become louder with harder strikes, as we all know, but I'm not sure I've seen a study about the relative amplitude of partials at low strikes and at hard strikes.  If a soft strike produces, among many, many partials, a fundamental and a 4th partial that are, say, at a ratio of 6:1 in amplitude (a wildly chosen ratio), does that ratio remain constant on a hard strike? I imagine that the answer is, roughly, yes. In any event, the hard strikes in Pianoteq sound very realistic to me. 

But of course, regardless of what the reality is, I wanted to try the experiment, both to hear the sound with partials having a different ratio to velocity and to see what kind of strangeness I could create.   (What would happen if the Q-factor increased with velocity, for example? If the Soundboard Cutoff was modulated with a subtle LFO? If the Strike point varied a little with hard strikes, or if the Stereo width parameter on the mics page increased slightly with harder strikes? The electric pianos might offer more fun with modulations.) On the other hand, the possibilities are a little frightening.

I've had some time to think about this situation. My impression is probably the same as yours--that the construction of the program and the demands on the cpu prevent modulating some of the parameters with velocity. Of course, some of the variables, such as the hammer hardness, are already modulated by velocity. But for the others, when we move a slider, we change variables--numbers--in tables that are referenced in the long non-linear equation that is Pianoteq. When we play, the sound is produced based on those variables. The cpu just can't keep up with the thousands of calculations that would have to be made if the variable changed with velocity, since almost every other parameter would be affected. In addition, if several of the parameters were modulated by velocity, which then changed the force applied in real time playing to other parameters which are themselves already modulated by velocity, I would imagine that the calculations and cpu load would be tremendous. The situation with a sample-based player is much simpler. Each parameter is isolated, so that a filter can be modulated with velocity, for example, without requiring new calculations for the amplitude envelope.

It's a pity if my guess is correct. I would love to be able to modulate the unison detuning inversely with velocity, or the reverb mix. And like Eran, to map the loudness of a given partial to velocity.

But I do sometimes wonder if some of the parameters could be coded so that they could be mapped this way to respond in real-time without causing a crash. Some might almost guarantee a crash when a few notes are played--such as the soundboard impedance. Some might be easier than others--such as the amplitude of single partials. Would require a lot of trial and error and guesswork. Philippe and Julien?