I've been working on an implementation of the tessendorf paper on ocean surfaces. I'm pretty well along, but I'm realizing that with per-vertex normals, the specular reflections I'm getting will always be pretty blurry. Like what Keith Lantz has going on in his pictures vs. Eric Bruneton's glittering BRDF. A big part of that is that Bruneton is generating a screen-space grid with per-pixel normals, while Keith and I are generating 3 dimensional grids with per-vertex normals.
Anyone have any experience with ocean rendering? What's my next move here, should I just scroll a …