I've been messing with the loop functionality in Substance more lately so I've decided to improve upon the default vanilla RTAO node. I've overhauled the majority of the core logic (specifically the ray march and mip-max stepping) while optimizing where I could.
Notably, I've moved all of the positional data inside the pixel processor so the input height does not need to be 32bit (keeping down compute cost).
That being said, the main difference between RTAO and JW RTAO is, when scattering points, the hemisphere of possible ray directions is not oriented to the surface (like the vanilla substance RTAO), instead it uses a fixed hemisphere direction oriented towards "the camera". This resolves the common "fresnel" look that appears when using the vanilla Substance RTAO.
I hope this helps for anyone who likes the visual result of HBAO but wants the per-pixel accuracy you come to expect from software ray traced solutions.
Download the sbsar here for FREE: https://artstn.co/m/klrBg