Finally did some tests with per-fragment panning!
At the moment I am not entirely sure how effective this is sound-wise. The panning movements within the individual fragments can get easily lost when many voices are playing at once or when the fragments are short etc…My initial hunch that the static pan positions per fragment are enough for HourGlass might be right.
The additional CPU load might not be too much though and the per-fragment panning can also be turned off. For now this is just doing a simple linear interpolation (straight line) between 2 pan coordinates. Allowing more complicated paths might be more interesting sonically, or not…
Making a suitable GUI for this may be quite involved. The solution which seems obvious at first, just letting the user draw a pan path is not really enough. Each fragment should be able to have its own panning path. But the user can’t really be expected to draw hundreds of paths etc…Now, there are solutions to this like the user drawing a few paths which are then assigned to the voices in a round robin fashion or randomly etc. But I need to think this further before making anything too complicated since it’s not entirely clear if the per-fragment panning paths even sound that interesting at the moment.