UT3 Map WIP thread
polycounter lvl 18
If I make a thread, I have to work on this! That is the idea.
Over the holiday I start mucking around with this layout. I made a bunch of DM maps back in the day (quake 2). All of my favorites were the small, 2-4 player levels. This will be the target size of my level.
Then filling out some the idea with a few materials and props (yucky wood and floor)
Today I spent some time (~1hr) on the archway collar. This is a material test on a simple mesh. The final will be much more detailed. The high poly is textured so all I have to do is bake it all back down.
My biggest hurdle is that my current PC can't handle UT3. I may install it at work and assemble there, while propping and material creation from home. I'll be using Max, Photoshop and Crazybump for most everything.
Over the holiday I start mucking around with this layout. I made a bunch of DM maps back in the day (quake 2). All of my favorites were the small, 2-4 player levels. This will be the target size of my level.
Then filling out some the idea with a few materials and props (yucky wood and floor)
Today I spent some time (~1hr) on the archway collar. This is a material test on a simple mesh. The final will be much more detailed. The high poly is textured so all I have to do is bake it all back down.
My biggest hurdle is that my current PC can't handle UT3. I may install it at work and assemble there, while propping and material creation from home. I'll be using Max, Photoshop and Crazybump for most everything.
Replies
Also, buy a new PC you cheapo!
I am very curious to see the final level design layout.
Small size seems more than reasonable here since the effort needed to create stuff for UT3 is way bigger that for previous releases
I am working on my own DM map now, so I'll also post some WIP later on. Meanwhile I'll be watching ur progress.
Btw, I think that for ur portal it would be sufficient just to get normal maps through Crazybump and not to model hi-res version.
Matroskin: the reason I create high detail meshes is because my technique is fast. I've been making texture like this since 1998*. It's really no problem to bang out a high detail texture mesh and bake it for normals. Additionally, I prefer the way it looks (the punch, depth, and no overlays).
*And those crazy bastards kept all those high detail texture models to bake normals for a next gen game here recently! And by crazy, I mean smart.
Also, how did you get such deep definition with Crazybump without getting so much noise? I find that if I want to get deep grooves I always end up with so much extra bump. What kind of settings produces such clean textures?
Love the ideas so far, seems simple enough to get done considering the potential length of a custom UE3 map. I'm inspired to come up with a layout for a small level myself!
Johny: Funny thing, I actually went into the blue channel and REDUCED the contrast of the texture (so I can revert back easy). I felt as a floor, it should appear flatter and less "I'm going to trip over every stone". But again, as stated above, if it sucks in game it's going to get the treatment.
Saidin311
As stated, the arch is created from a high detail mesh, but because you're so sexy, I'll go into a little more detail...
As far as reference goes, yes, I googled "brick arch" to get some ideas how I wanted the bricks to be laid out in a proper arch fashion. But that's as far as my reference went. I reference almost everything I do, usually creating a collage director of images to fall back on. Better to do it right that spend a ton of time on something only to have someone reply "impossible, retard!"
I start by modeling a brick, and cutting unique chunks/edges out all around it (this is important for making it appear like a unique brick from different angles). Then I instance and rotate/scale them all around to make my whole arch. Even the keystone is the same brick, but with a few more cuts (the 3 base stones are different as their size/position demanded it). Then a subtle noise over it all to make everything appear even more unique. I apply a diffuse to those bricks, as well as mortar mesh underneath it all. It's really a very easy process only taking about 30-45 minutes. It takes about the same amount of time as cleaning up a picture of bricks, and yield those hard, clean edged to everything which is a look I prefer.
I didn't actually paint anything except for some minor color/roughing up work on the high detail diffuse which is placed on the bricks/mortar. I don't put much thought into this, just a lot of color variation.
Final High detail mesh
Then a make a low poly mesh to bake the diffuse, normal and lighting map (AO) to a texture. I put the diffuse this generates into crazybump to pick up the small rock/crack details and overlay it on my normal generated from the high detail in max (reducing the blue levels to 128 so it overlays properly). And I multiply the lighting (AO) over the diffuse to add depth the D map.
Now, on the floor texture above, I also took the high mesh into Mudbox for a quick muck-up, as I felt the max work wasn't enough. I did the same with the brick wall in the background of the book case scene. Again, I don't spent much time in there, just quickly roughing things up.
This is a technique I've used for years, and thankfully, mikebart took the time to document his version of basically the same process. His results are excellent.
http://www.doom3world.org/phpbb2/viewtopic.php?p=160978
im guessing that'll come in the nect iteration when your happy with overalll?
also floor lacks for having blue channel played with, love the subtle specular shines though
You ought to try taking your hipoly as it is, then merge it
instead of submerging each subdivided brick into the core mesh. So that its all a solid connected mesh or as near as.
Then you can take it into Zbrush, still rely on the symmetry modifier
in there so you can retain some texture optmization later when processing.
Anyway, if you do merge then zb and subdivide a few levels
you can paint the erosion of the bricks into the grout
and have it line up correctly so that when you split off
>500,000 (250,000 in zb!) chunks to export to obj and use
the cheap polycruncher addon in max to optimize by 70%.
You can also split off a version of your ZB model and lower
the entire object to the half a million poly mark and export that as a whole back to max.
Then you can polycrunch that solid half a mill mesh down to your target polycount for the lowpoly and it will retain
a little bit of extra articulate in the right spots so that
when you process it will truly seem more next gen and support the deformations you made to the expensive version you needed to export in chunks.
At least, this is generally the way I work now and it saves building the low completly by hand and it avoids the sharp edges that trying to optimize too much by hand leads to.
Other than that its looking good, just though I'd suggest something extra to experiment with.
k.
Until next time, here's some stairs. Again, quick low poly, you can see from the silhouette it's just box stairs, nothing fancy.
Until next time, here's some stairs. Again, quick low poly, you can see from the silhouette it's just box stairs, nothing fancy.
[/ QUOTE ]
If you'd allow me to make a comment on your stairs I would! The clear definition in the stones on the 2nd, 7, 8 10 and 11th stairs takes away from the fact that there is no break in the stone on the 1, 3, 4, 6 and 9th stairs. Somehow my eye expects to see at least some break in ALL stones, as if I never had a full sized flag stone to construct one entire step and had to settle with broken stones/sand/mortar for all steps.
http://img2.timeinc.net/toh/i/a/design/carlisle-house-30.jpg
http://boards.polycount.net/showflat.php?Cat=0&Number=261755
But seriously, it's all for variety sake. Could you enjoy those breaks the same if they were always there? Two grand areas aren't as grand unless they are connected by a small area to give a sense of scale. In the words of Butthead "If nothing ever sucked, how would you know when it was cool?"
Last night, Brome starts massaging me. Telling me I need to show my steps for everything cause I'm a pro and we pros need to help out when we can. So here's my steps for my steps
1. Layout of the design I want for my area (as seen in design pictured above)
2. Modeled a brick with some cuts
3. Laid out the bricks, offset a few, smooth them.
4. Added a dirt layer mesh to show where dirt would build up on the steps. This was as simple as taking the mesh from step 1, smoothing it, add noise and manually correct where necessary.
5. Took the steps into mudbox and quicked some rocky lines across the front, and pinched the top angle (where a step is the sharpest)
6. Back in max, applied a diffuse stone to steps and dirt to dirt mesh.
After that its baking materials and applying them back onto step 1 for the final. That said, when I do complete the FINAL IN GAME LOW POLY MESH, I'll chamfer the inner dirt areas, add a few extra cuts/edges to the mesh and shake it up a bit to create imperfections.
Too many people rely on crazybump for enviornment normals work
the engine doubles the vertex count when 2 smoothing groups
are next to each other to get around the fact that the engine
doesnt use smoothing groups.
An added bonus of using 1 smoothing group is that is makes for larger amounts of colour in the normals which forces more depth ingame.
k.
Thanks for the tutorial.
how do you deal with the smoothing errors that occurs on the 90 degree angles then?
I'm getting this stuff all over though when dealing with hard surfaces and I have to split my UV's and smoothing groups to fix it.
the way to see what you are really getting with different processing methods and smoothing groups you experiment with
is to place it in a level.
I generally fake, fudge or manipulate things as much as I can by trying to take advantage of the issues and strength's
of the engine, such that I am aware of anyway.
For instance, I often force 1 smoothing group to do too much work and experience of doing this in certain cases
has taught me not to worry about the GB or max previews
because sometimes forcing 1 smoothing group to bend over too
extreme an angle persuades the engine to do the same and I
get more pop from the normals than I would have if I used
multiple smoothing groups or chamfers to control a single
smoothing group so its more appealing when viewed on its own
in max in a vacuum from the notion that it will carry a normal map later.
I've tried to explain previously that the lowpoly mesh is better thought of as putty now, placing a putty like low
poly in a level and rebuilding the lighting is like hardening the mesh.
I've sometimes unwrapped stairs like this and when thats
finalized I add lots of edge loops on the flat faces so that
I control the smoothing group the way I would with chamfering for proper articulate.
Then I process, take the edgeloops back out and put it ingame and see if its helped me, again, fake or fudge the
look of there being more there.
At least thats what I try to do in cases where its a very primitive form thats being processed. If you have UT3
you can check out the concrete Liandri (LI_)bunker walls
that are used in all the cold snow levels, they all carried
hideously warped overextended smoothing groups and it made
them look far better in the finished levels than being
conservative and following the logical rules of smoothing
group use yielded.
k.
here's an example I made:
the three models all have exactly the same geometry. However the two on the right have 1 smooth group and the one on the bottom left has 2 smooth groups.
The way I made these.
1) first I made the basic steps you see in the bottom left
2) then I set all the smooth groups to be 1 to make the model on the bottom right
3) then I created a normal map using the model on the bottom left as the hipoly and the bottom right as the lowpoly.
4) I then copied the model on the bottom right and applied the normal map to it with a directX shader. This gave it normals that looked like it has smooth groups when in reality there's only one smooth group.
Having done lots of smooth surfaces on objects that need to be dynamically lit (vehicles) I've found this issue to be extremely time consuming to deal with.
*I'll end this de-railing here and take my issues where they belong
If you put a normal map from a hipoly source on that it will usually fix the smoothing errors. Sometimes you will need to chamfer the edge if it's sharper than 90 degrees
here's an example I made:
the three models all have exactly the same geometry. However the two on the right have 1 smooth group and the one on the bottom left has 2 smooth groups.
The way I made these.
1) first I made the basic steps you see in the bottom left
2) then I set all the smooth groups to be 1 to make the model on the bottom right
3) then I created a normal map using the model on the bottom left as the hipoly and the bottom right as the lowpoly.
4) I then copied the model on the bottom right and applied the normal map to it with a directX shader. This gave it normals that looked like it has smooth groups when in reality there's only one smooth group.
[/ QUOTE ]
This example here has smoothing errors man, so you're just proving why you dont want to do this. Adding in a few extra edges is definately the way to go, esp. on an asset thats only like what, 20 polys? You could add a LOT more detail to that before you'de have any performance difference, at that point it is likely going to be the same hit to render 20 polys, or 200 polys. The hit you're taking here is from the amount of pixels actually being drawn on screen, because you're nowhere near causing any problems with amount of tris on screen.
Though, something i noticed, at least within 3ds max, if you have normal maps applied to the materials in the high poly, they will carry over to the low poly along with the high poly normal mapping. Which takes out the need for overlaying normal maps, unless i am missing something.
http://img61.imageshack.us/img61/5638/choldenfireus1.jpg