Looks nice. I would just slow it down or filter in the time domain. Anything else will change the overall appearance and that's a subjective judgement - do you like more blue?
The key was to calculate everything on a base of 256 fractions between the integer steps.
Now I need a better color-/gamma correction which allows that 2 leds at 50% are exactly as bright perceived as one led at 100% brightness.
This might become tricky due to a different human sensitivity to different colors and the different led characteristics itself.
I´ve no idea yet how to start with a decent calibration. It might need a couple of test pictures with gradients in order to bend the curves into the right S-shapes.