UE5 Procedural Grass Using Bezier Curves pt. 2: Calculating Normals

[@minidinogamer]

This is such a good tutorial! I skimmed through all the videos and i'm so glad with the results you showed us. I've been having trouble trying to optimise my own grass but I cant get the amount of blades I want the player to see, unlike your end results, so this is amazing. I am having an issue with the normals though, their is no gradient between the normals, it's just two colours which is the cream colour and the blue colour, and in lit perspective it's the base colour and a much darker colour of the base. My nodes for the normals are the exact same as yours so i'm guessing it's a material setting? I'm very new to materials though so I have no idea why my normals are different from yours. Any help would be extremely appreciated!

[@TechArtAlex]

I don't think there are any material settings that would cause that, but it's hard to say. I'd start by triple checking the nodes. It can be helpful to preview each one to see where something suddenly looks unexpected. Also be sure to check out the intro to PCG compute shaders. It's not part of the series technically, but it has potentially big implications for improving the performance.

[@minidinogamer]

@@TechArtAlex Fixed it! all it took was re-doing it. I'll definitely check out the PCG compute shaders vid. Optimisation seems to be the biggest hurdle for me so far and im sure that vid will help me out

[@dcrookz]

this is very cool. I would be interested in seeing more on stuff from the tsushima foliage talks. IMO they have the best foliage in games... at least as the apex of tech vs art-direction. Alan Wake might have had more technically complex foliage but Ghosts looks the crispiest. I would be interested in particular if the mesh provided was say, several planes that were procedurally spread out, rotated and projected down onto a landscape (or other surface). I feel like you could massively reduce the instance count of a field of grass but preserve the random rotation per blade look. This would probably require storing some data in the mesh... I'd also be interested in seeing you procedurally adjust the normals so that the blade looks lightly convex/concave along the width of the grass, with rounding as it gets to the edges. But you know, I'm just one guy on the internet.

[@TechArtAlex]

Thanks! I should be able to do all that. I think the instance count count with nanite is by default limited to about 2 million, so having multiple blades per patch could definitely be useful. It wouldn't be too difficult but it would mean you couldn't rely on the mesh transforms for length or rotation, so all 3 control points would need to be positioned in the shader for each blade. Each blade can be identified easily with vertex color or UVs. I also thought it could be interesting to combine multiple blades into one single taller grass plant, as many grasses have multiple blades along a stem. Adjusting the pixel normals for some added depth is surprisingly easy so I can definitely show that.

[@ScorpyX]

Amazing

[@TechArtAlex]

Thanks!

[@unlastgamer7979]

When I turn on world normal my grass turns into rectangle i rechecked everything but nothing seems to work. Any idea how to fix it?

[@TechArtAlex]

Do you mean when using world space normal instead of tangent space normal maps?

[@unlastgamer7979]

@@TechArtAlexNevermind I restarted my pc it’s working fine now.

[@oliverbendsen]

not sure my normals are correct. everything is the same as yours

[@TechArtAlex]

What makes you say they are wrong? It can help to bring in a cylinder, sphere or some test geometry, and then look at the buffer visualization viewmodes for the world normal. That way you can compare it to something with known correct normals. I've already validated mine as correct, so if yours aren't you may want to review everything again.

[@oliverbendsen]

@@TechArtAlex i fixed it. had p2 and p1 switched mb. absolutely amazing series! im on the last video now. i hope the next video is optimization

[@TechArtAlex]

@@oliverbendsen Awesome. Thanks! The next one should be optimizing and larger fields, although there may be multiple parts on optimizing depending on how long the video ends up.