Yes, it is difficult, definitely... with no definite answer. It's something of a compromise either way: On the one hand, a robot playing the key doesn't sound right to me, but it has more accurate velocity control; on the other hand, a human playing the key has the right touch but the velocity control is not so accurate.Lode_Runner wrote:The reason I felt that a machine would do a better job than fingers for the sound of the hammer striking the string? When you're working with as many as 25 velocity layers, I think that it would be very difficult to accurately physically play a key at 25 different velocities with human touch.
For me, a big problem with even the best computer controlled robotic actions is that nothing actually strikes the key. This leaves staccato (and other articulations where the finger strikes from a distance) lacking in realism.
If we had unlimited time and unlimited hard-drive space, we'd have a human playing the samples and during the sampling sessions, record everything in 100 velocity layers. For that, we'd choose 16 for the end product.
The producer or keyboard player know that this nuances do nothing to the final recording and don't really care; they use to listen to the piano in a concert all or in a live jazz/pop gig; at the same time the pianist can't be inspired playing such cold sampled instrument and lend toward an acoustic recording that make him happy while the engineer get crazy trying to cancel the awful 'noises' that pianist like so much but reduce power and clarity into the mix. The keyboard player gets a little confused but doesn't care very much, if it works live or cuts in the mix is ok.. argh.. 
