Grass edge 4: Vergemaker

[nbohr1more]

Have you guys been in contact with these folks?

 

Doom 3 Mega-Texture Mod V2

 

From what I can tell you might need ETQW to author the megatexture but I was reading about some other methods that seemed to do away with that requirement (reading comprehension was low as usual )

[Bikerdude]

Looks like Rich_is_bored has been taking a look...

[Tels]

Looks like Rich_is_bored has been taking a look...

 

Well, the mod says:

 

======================
COMPARISONS TO QUAKE WARS
======================
* Terrain scale is ~2/3 of Quake Wars maps.
* The 32768x32768 Megatextures in Quake Wars has 128 times the texture
detail you have in this demo.
* What would be a 5.7GB Megatexture in D3/Q4/Prey takes only ~500MB in
Quake Wars. Or in other words: The 2048x2048 Megatexture in this mod
would have been 2MB and not 22MB.

 

So it is less resolution and much higher memory and bandwidth. And nearly not as good looking (but that might be because it is a demo).

 

Something we should try, nevertheless.

[Fidcal]

Experiments in a real map with both horizontal and vertical went well. This image shows both and you can see I let the vertical cross right over a doorway - mapper's choice but see later. In this crouching close up shot the horizontal fade is a flat big fuzzy edge whereas the vertical is a bevel with the trim edge.

 

 

Here's another crouching closeup of the vertical and is a fair representation of how it is in game. If anything it is marginally better in game.

 

 

This is a standing shot and it clearly removes the sharp edge between lawn and wall that you can still see on the untreated step at left.

 

 

This is how it looks in the editor. Later I'll do proper editor images. You can see the horizontal one on the right has the big fuzzy edge sprawling onto the stone alcove. It uses the trim edge to blend with the rest of the lawn. That trim edge seems probably good enough. I was unable to positively find it in game even though I know where it is but that might depend on alignment and texture type so needs more testing. If OK then only one strip is needed on the tile and considering the amount of pure white and black probaly the tile can be reduced to 4096 x 2048. There seems to be quite a margin for error but I want to do more testing with the tiles deliberately misaligned to the main lawn to see if the join is evident.

 

 

This final shot shows the back of the bevel from inside the doorway. I see a black band where I would expect transparency. Any suggestions? Mmmm... I thought I had given this the decal keyword so maybe that will fix it. Two-sided might be considered to. I can feel the bevel underfoot when I walk through the door so maybe it wants to be a func_static solid 0 (or I think decal in the shader does the same.)

 

 

This also raises the question where these shaders should go in the hierarchy. I see them as decals but should they go with their 'parent'? I expect to see at least three grasses, probably straw. Working title in this test was grass_short_patchy01_edge.

 

Summary: this is very workable and although it needs practice how best for the mapper to apply it, it is not too difficult and reasonably flexible.

[Fidcal]

I just rotated the main lawn texture 45 degrees to the edge patches and I still cannot find the join in-game so that is very encouraging. It suggest no need to even bother aligning but just make sure the scale is approx the same.

[SneaksieDave]

Wow!

[Bikerdude]

looking superb young man!

[Fidcal]

I can't get this to work with a cropped 4096 x 2048. If I make the patch the same texture scale as the rest of the lawn then the fuzzy area blades of grass are twice the size because they are a product of the alpha map. If instead I make the patch half the texture scale then the fuzzy blades are correct but the texture is compressed.

 

If I put scale 1,0.5 on both the diffuse and bump stages then the scale works but the bump is heightened compared to the rest of the lawn!

[Fidcal]

I gave up on the 4096 x 2048. It's actually easier to work with a square with more blank alpha because you don't have to be so precise. With 4096 x 2048 there is only a narrow edge of total black so you have to make your patch an almost exact texture tile size.

 

Anyway, I've made a load of these in one place and made them func_static and using a lever to hide/unhide them. I see a big performance hit. In the worst case from 30fps down to 20fps. I presume this is because of the size and number of them.

 

An extra problem has arisen because all of these materials tile only one way - along the length of the grass edge. This is fine if both ends stop at a vertical wall but otherwise there is a straight of grass overlaying grass and it is visible. So I have had to make a second patch which is a square with fuzzy edge on all four sides. This can then be laid over any such lines to hide them.

 

Anyway, I have now reduced the alpha tile to 1024 x 1024 with very little loss. 4096 was derived from my supercrumb but this is not supercrumb but grass so it seems OK at 1024. So now the performance hit is lower. In one position at 24fps it dropped to 22fps. In addition, I'm hoping this is because the func_static surrounds me. I suspect if I make them all separate one then the ones behind me won't be processed.

 

[EDIT] Also I hadn't yet converted the square patch to 1024 and also this test was a lot of patches in a small area. So overall I am hopeful the performance hit will be tolerable.

 

One other thing. I have set this up so the mapper can set the thickness of grass edge on each func_static using shaderparm3 0 to 1. I'm checking this won't conflict with LOD but I think that unlikely. What this means is that apart from setting a general thickness easily to suit the mapper's taste, it can also be varied so if you have a neat lawn in one area and some scruffy grass in a corner you can vary it. In fact the extra fuzzy stuff looks more like moss when against a wall so it can go up quite a way.

[Bikerdude]

I look forward to playing with it Fids..

[Fidcal]

Sorry this is taking so long but there are so many elements to it to think about and I get burned out after an hour or two so it's spread over weeks. But I'm gradually getting there. Actual placement of the patches can be fairly quick once I've pinned down the do's and don'ts. Current philosophy is a 1024 x 1024 but with less full white (opaque) so there is more full black (invisible) so you don't have to be so accurate in patch size.

 

I've virtually done Biker's yard but need to finalize all the materials, editor images and probably make some prefabs and tutorial. Biker: you probably will have to place the odd weed to cover tiny gaps eg, where the bevel ends at the doorway. Although it might be possible to add a column to that patch and bend it round. I tried an endcap but it was crap.

 

This is another path in my own FM. I considered whether to make a model. Both methods have their merits. In this case I though it would be another good test of the new alpha material. Basically I made a shallow caulk pit. Then a hard dirt patch in it sucked down in the middle to form a curve. In the DR shot you can see I laid two patches with the grass blend curved down from either side into the path. In this case I use the fuzzy edge on the path and the grass blades edge to blend with the rest of the grass. I make them func_static and give them shaderparm 0.5. Now I can tweak that if I want to make the grass edge advance further or lesser without having to move anything or dmap. It's not perfect but I'm very happy with the result.

 

[Tels]

It's not perfect but I'm very happy with the result.

 

And you should, the second pic looks very good (only slightly than a real vertex blend).

[Bikerdude]

looking good fids.

[Baddcog]

Yeah, it looks great.

 

I think perfection on this is almost impossible, too many variables that can change map to map, author to author anyway.

 

And even in Hammer where you can vert paint in editor (which is very nice) you still have the same issues, aligning textures, transitioning from one blend texture to another...

[Fidcal]

Yes, the above path curves around in the foreground in a rounded L shape. I've just about got it to my satisfaction. Some parts were fiddly but then a lot of patchwork is I guess.

[Fidcal]

Nice - this works in reverse by setting the dirt or stone in the material. Hey - instant path! Just throw it down on any grass anywhere. At least with a flat path you can. This quick test is rough and wants refining but no alignment is needed. You can move this func_static path around like a model but can retexture it, lengthen it, bend it. It has to be a texture tile width though so the width will depend on what scale you use. So if you are not too picky you can reduce or enlarge the width if you don't mind the scale being slightly different. It does not have to match the grass scale or rotation or anything. Oh and each far end has to be patched too unless it hits a wall like ends at a doorstep.

 

I'll make some materials from the common dirt and stone floors and some ready to go prefabs.

 

[Bikerdude]

hows goes the grass with my map Fids..?

[Fidcal]

I can probably upload it this weekend, maybe even today if you want. I am trying to bring everything to completion together as a standard set of materials. But there is probably no good reason to wait for that. Your map is using a couple of materials that might be slightly different in the final version but that won't affect your map which will be independent.

 

I'll look at that next to give it one more checkover. Then I have to package up the materials and upload it as a prefab to import into your map. You might need to raise your moss patches a tiny bit just above to avoid z-fighting.

[Bikerdude]

no rush, as i wont be playing with that map till after the contest..

[Fidcal]

Well, you can still work on it. These patches will just import into your latest version.

 

Just checked back and having to tweak a few things. I had it OK before but somehow it is different - probably I changed the materials externally. Just that big mound at the end which is difficult.

[Bikerdude]

Just that big mound at the end which is difficult.

get ride of it for the time being then fids..

[Fidcal]

I deleted my patch around there because I'd tweaked it so much it was like a screwed up newspaper and did it again and it was easier the second time.

 

I have it all in a second folder and imported the prefab and it works on a clean copy of the map so I'm nearly ready to upload it.

[Fidcal]

OK Biker, you can get the materials and prefab for your yard from http://www.fidcal.com/TEMP/YardGrassEdges.zip

 

NOTES:

 

After looking around, frob the lever by the pond to see how it was before. When done you can just delete the test lever.

 

Your original moss patches will need raising up to avoid being buried and z-fighting.

 

The doorway grass will look better if you push the door further in I think.

 

One or two small spots you might want to cover with a weed or something.

 

You can tweak the depth of the fuzzyness on each patch entity by changing the shaderparm3 value 0 to 1. Low is longer, high is shorter.

 

I've not done proper editor images for this so it just uses the alpha black and white as you've seen in my earlier image above.

 

I'll be preparing a more advanced and varied pack for many materials and prefabs over the next few days or so and upload it for everyone with a tutorial.

[Bikerdude]

Smashing, thanks Fids - well done that man! the grass looks spot on I cant even find those spots that needed covering, it all looks good to me.<br>

Edited August 14, 2010 by Bikerdude

[demagogue]

Oh and each far end has to be patched too unless it hits a wall like ends at a doorstep.

 

So are you also making path-ends? It might look good if you had a patch of the path ending in either a fading edge or fading rounded-edge. It's great you thought of doing it for the path, since that will be so much more versatile.