2 and 4 were easily done for me. Mathwise, I can’t seem to make a cheaper heightblend vs what comes stock in the engine, but it’s performant enough for my needs and I only call once per layer used, thus #3 is a non-issue for me.
#1 I’ve tried to work with but I don’t seem to be able to wrap my head around it. I kind of get what you mean by scaling the UVs based on distance and controlling what you see that way (which is really ideal in almost all cases as the UV math will be cheaper and further upstream), but in practicality, I can’t seem to make it work.
Would you have a direction or example on how this works? I always seem to get a seam. I then had the thought to use 2 samples and offset the distance-math so the transition of one was hidden by maximally blending the other, eg: for that one/few pixels where the UVs change for sample1, those pixels would only be sample2, which blending on distance away from the camera would mix back to 50/50 and then to blend out sample2 vs sample1, rinse/repeat over distance and increasing levels of scale in the UV math.
Sounds simple and elegant, right, but I couldn’t figure it out.
That aside, it seems like you will still need 2 samples/samplers to make it work, is this a correct conclusion? If I already have an effective solution based on 2 samples per-layer (micro/macro w/some decent blending), does what I described above gain anything (assuming I’m on target…)?