Large areas of grass - current techniques

[tricky]

We are creating a number of images in addition to animation for a large business park. 20-odd buildings and carparks, gardens, road systems, planting, trees and GRASS!

We've used various techniques over the years, from flat textures which look pretty awful, to displacement (which I never achieved good results with), to vray hair/fur and most recently ForestPro and some small tufts of grass from the iGrass library. The latter option probably gives us the best results, but it can bring our systems down to a crawl (no slouch machines either with 16GB Ram and i7 processors). Our farm has much less RAM so they struggle a bit. To get a good coverage over an area, this approach can require 3+ million copies of the individual tufts, randomly rotated and scattered. ForestPro is a stunning bit of software, but I think this method pushes it (and normal machines) beyond their limits.

Have people had good results with displacement, or any other methods? Care to suggest?

[super gnu]

ive had good success doing "infinite" grass using forestpro.. if you design the source mesh carefully (igrass is a bit heavy, and a few polys in the source object can make a very big difference when multiplied a million times..) and adjust the density to be acceptable without excessive, enough for, say 2-500metres of grass from camera, you can map the z scale to camera distance (and put the axis of the mesh below 0).


this causes the grass geometry to sink gradually below the landscape over a set distance from the camera. after its below the surface, you can clip it using camera clipping. get the ranges correct, and the map it blends into properly colour corrected, and you can fly down from space and have grass wherever the camera goes.

actually i believe the new version of forest allows you to control z-offset by camera distance, or use vraydistance tex, so the z-scale trick might not be necessary..)

ive used this to do flyovers of very large terrains ( 50-100 sq km) use splines to define the areas you want grass, then set-and-forget. whenever camera comes close enough, grass starts to appear very subtly.


this is my solution for animation at least. images can be comped much more easily.

[tricky]

you can map the z scale to camera distance (and put the axis of the mesh below 0).


actually i believe the new version of forest allows you to control z-offset by camera distance, or use vraydistance tex, so the z-scale trick might not be necessary

Good suggestion. Will look at that. Can't see the z-offset by distance, but will double check.

[tricky]

I can't even find how to map z scale to camera distance - can you point me to that?

[super gnu]

sorry, one sec, i was telling you off the top of my head.. i actually used a vray distance tex to control z-scale, and it was a bit fiddly to set up, but ive definitely read that the later version fixes this to work well, and adds options to map the other offsets.

let me fire up max and have a looksy.

[super gnu]

ok seems i was talking out of my a** about the new version.. they did indeed fix the vraydistance tex issue ( consider a falloff map set to distance blend might work more simply, if its also fixed - didnt work with old version)

however you still cant map the position offset with a texture (according to release notes) only rotation and scale.. so my original method is still the only option.. z scale mapped, with the axis a way below the geometry, so it moves down as well as scaling.

the use camera clipping to delete grass beyond falloff range.


youll have to make the landscape map get darker over the same falloff range as the grass shrinks, to compensate for the loss of AO and shadowing you get as the 3d grass dissapears.

[rivoli]

in forest you can either clip the camera plane, or use the density falloff. both are under the camera rollout. as super gnu pointed out, once you have the grass geometry optimized enough, you can render massive fields of grass..

[super gnu]

the problem with using density falloff, is you get ugly artifacts in animation, as random clumps of grass dissapear.. this only works really if the clumps are much smaller than a pixel by the time the density falloff kicks in.

+1 for being able to map the position transforms of the forest object! the scale trick is a bit of a hack.

[chrisjacks21]

ok seems i was talking out of my a** about the new version.. they did indeed fix the vraydistance tex issue ( consider a falloff map set to distance blend might work more simply, if its also fixed - didnt work with old version)

however you still cant map the position offset with a texture (according to release notes) only rotation and scale.. so my original method is still the only option.. z scale mapped, with the axis a way below the geometry, so it moves down as well as scaling.

the use camera clipping to delete grass beyond falloff range.


youll have to make the landscape map get darker over the same falloff range as the grass shrinks, to compensate for the loss of AO and shadowing you get as the 3d grass dissapears.

Hey Robin, nice technique - have you used this method and camera clipping with animations using pre-calc irradiance maps - do you not get issues where the geometry disappears?

[super gnu]

for sure it wont be 100%b correct, but in my experience, not really a problem, as the geometric changes are always very distant and subtle...

[tricky]

I'm going round in fecking circles here!!!

I have a number of iGrass clumps (yet to be optimised) in my FP object. The distribution map is set to FULL and the size is 8m. The pivot point of each small clump is 1m below their centres.

In the transform settings, I have scale enabled, with XYZ aspect ratio locked together. The min value I have as 1 and the max value as 100. I then have a VrayDistanceTex (though I have tried a falloff using Distance) in the map slot set to 30m. The default black and white swatches are in the colours of the near and far distances, though I have tried reversing these.

What am I missing?


Image attached now. Seems to be scaling from the centre of the 50mx50m spline, not the camera z-distance

[super gnu]

hmm what do you have the distance tex linked to? i used a dummy sphere, hidden from the camera, but linked to its motion. also id advise only scaling on z axis (not x and y) as the fade out stays more natural.

[tricky]

hmm what do you have the distance tex linked to? i used a dummy sphere, hidden from the camera, but linked to its motion. also id advise only scaling on z axis (not x and y) as the fade out stays more natural.

I've just created a non-renderable sphere and linked this to the camera. The VRayDistanceTex now has this sphere as the VRayDistanceTex object, black in the far colour, white in the near colour and inside solid colour ticked and set to white. Now I seem to be able to control the z-scaling by changing the distance value in the VrayEdgeText. Is this correct? Seems a little long-winded, but I think it is working.

<EDIT: so the size of the non-renderable sphere dictates the start of the scaling, and the VrayDistanceTex distance dictates the falloff...rrrrright?>

[rivoli]

the problem with using density falloff, is you get ugly artifacts in animation

aaa, ok. didn't know that, yours sounds like a good workaround then.

[super gnu]

yep.. its a bit long winded.. but so far its the only way ive found of using forest to fly over km of grass...


indeed.. radius of sphere controls start point, and distance tex controls fallof distance. takes some tweaking to get right thats for sure.