Concept to environment: Wasteland Waystation

GarageBay9

I finished hashing out a basic idea of the environment I'm going to build next, which is supposed to be a small dwelling and 'safe place to rest' in a vast desert. It's a couple ad-hoc structures built from found materials (shipping containers, scrap materials) shaded by a giant expanse of canvas that's stretched from the top of an old radio tower.

Here's the concept I have so far. Not the best painting in the world, but it's got the general feel and ideas I want to capture.



I'm making a breakdown of different assets I'll need to build for the scene (working in UDK). Any critiques or comments on the basic concept (or, heck, my painting skillz )?

GarageBay9

SasoChicken wrote: »

[Nerd warning]

This totally reminds me of that one episode in Star Trek Next Generation where Q is trying humanity and puts Picard and them in this desert and there's a setup just like this.

My wife could tell you all about that episode... but I'm not a trekkie, so I'm gonna plead ignorance.



(Now I have to bug her for a screenshot... )

tyl3r

I like the concept but feel it may be lacking a strong focal point right now. It's like a bunch of props thrown together in a desert. I would say maybe add a few more items, and try to find a strong focal point. Also, adding tears in the canvas I think would look pretty sweet!

Also, this really reminds me of some Killzone 2 art. Especially in Suljeva Desert or the MP map Pyrrhus Rise. Here's some screens to maybe get your juices flowing if you don't know what I'm talking about!

Screen 1
Screen 2
Screen 3
Screen 4

I'll definitely keep an eye on this thread though as I think it could turn into something really great!

GarageBay9

Ok, I'm going to shamelessly lift adam's scene breakdown format because dammit, he did it right.


What I'm Trying to Do:

- Produce a very competitive portfolio piece (this is the big one... I'm going on sixteen months unemployed and have to land some kind of job in my area very, very soon)
- Transition from U3 Editor to UDK
- Become much stronger at complex shader construction, which I suck at right now
- Accelerate my production speed
- Solidify my next-gen workflow, which is all over the map right now
- Build some skills and experience using next-gen tools to build natural environments with smooth transitions (MeshPaint, Terrain, vertex-color based shader transitions, etc)
- Accomplish all of this without completely pissing off everybody on Polycount with constant questions

Assets (2D)


Sand. Lots of sand.

(quick explanation: the area under the tarp, even though there are no walls, is kind of considered 'inside' since it's the sheltered space. Everything that isn't under the shade of the giant tarp is 'outside')

- Undisturbed sand
- Smooth sand in lees and sheltered areas
- 'domestic' sand (sand on floor of Conex containers, etc)
- Disturbed sand
- 'Wild' sand
- 'outside' blown sand
- backslope of dune sand
- 'Inside' sand
- 'under-tarp' blown sand
- frequently walked in sand
- churned up sand in work area

There's no water and no vegetation so I get away easy in that department... Trying to avoid cacti or other particularly iconic desert plants, want this place to look almost like the planet in Pitch Black.

Assets (3D)


'Anchor' pieces

These are objects that create the backbone of the scene and the environment.

- Broken antenna base
- Broken antenna top (have ref for both)
- Massive shade tarp (have some ref, have to fudge the rest, nothing this big really exists)
- various steel grating for artificial floors (have references)
- Cables for tarp
- anchors for tarp cables (have ref)

Large scene objects

These are large pieces that the basic environment could still exist without

- water storage tanks
- water condensers
- primary solar panel array
- battery array
- sand breaks

'Inside' objects
- Radios
- Tables
- Solar panels
- Small antennas
- Conex containers (several sizes and colors)
- I beams
- misc. scrap metal sheets
- jersey barriers
- modular steel ground plates

'Garage area' objects
- tool boxes
- loose tools
- engine hoist
- wheeled work platform (just an ad-hoc mobile gantry)
- anti-sand clear tarps and stands
- anti-sand air filter

'Living area' objects
- wind chimes
- books
- canteens
- pot, pan, ladle, knife
- stack of canned food
- ladders up to hammock
- hammock
- ammo cans
- old rifle
- spear
- sand shoes (like snowshoes, but for sand)


That's what I've got for now. I'm going to start sketching quick concepts for some of this in photoshop and hopefully post a batch of those in a day or two. I'm also blocking out the core layout in Max, just getting scale, arrangement and volume down.

theStoff

Looks pretty awesome. HOw are you planning to make all the sand? Is that something you'd do in 3D or added in after?

GarageBay9

kosh_fotsirk wrote: »

Looks pretty awesome. HOw are you planning to make all the sand? Is that something you'd do in 3D or added in after?

Probably a little of both. That's one of bigger hurdles I'm working on. I'm thinking it'll be a mix of terrain, static meshes, and blending between the two with MeshPaint and vertex color-blending in shaders / terrain materials.

Here's a quick concept for the radio stack, which I'm thinking of doing first so I have a solid next-gen asset to shove into my portfolio in a hurry.

GarageBay9

...and here's a quick render of the high poly.

vcortis

Hey this is looking really good, I'm not sure how this is all going to bake down to the low poly with those switches, especially if you have them comming straight towards the camera and not flipped up or down.

GarageBay9

Thanks!

I'm going to bake just one and map all the low poly switches to one spot. They're mostly there in the high poly for positional reference... and to look cool.

Going to do that with most of the small details, at least for the normal map UVs. Some of the knobs will probably get unique diffuse UVs.

EDIT: Here's with some zbrushed dings and dents added. Now for low poly.

GarageBay9

Update with the Radio: here are shots in the Unreal engine. Lighting could probably use some minor tweaks, it's a little yellow.

Xoliul

Wow, 5800 tri's and a 1024 is huge for such a small asset. You better put that very close on the foreground to justify all that detail
It looks good though, maybe a bit uniform in colors.

SaferDan

yeah 5800 is loads for that! If i were you i would have normal mapped the whole of the back of it!

having said that it does look very pretty! The rust placement is epic!

vcortis

Oh god I just threw up in my mouth a little bit. It's almost 6k tris?!?!?!?! That's as much as a character in a lot of games. In my opinion the asset is just way too small to be that high. You really need to try to get it to maybe 800(which is still a lot for something that big) tris or less with a 512x512 texture.

I'm already seeing a lot of place you could cut tris easily without sacraficing a lot of detail. But a wireframe would be nice.

As far as texturing goes the diffuse looks pretty good, but the specular looks like it's overly simplified for a metal object. I'm not getting any of those light shallow scratches for an item that's in the desert in the highlighted areas.

Again texture flats and I can help. Looks like you're on the right track though, the highpoly turned out beautiful. But I'd go back in and add some of that metal corrosion to it, because right now the peeling paint is looking a bit flat.

Surfa

Jesus with those specs you are into first person weapon territory. Some of the little switches should be just normal mapped. Also you could tone down the number of sides on others.

Pedro Amorim

God! Leave the man alone. Let him do his thang. GOD

Shogun3d

Garagebay,

Nice work first of all, the high poly mesh looks great and your bake looks like it's doing its job very well. Your textures are also very nice although I feel you can push the specular a little bit more, particularly along the edges and chippings.

As bluntly stated above, depending on the engine, the tri-count is high and texture size is large (which isn't a problem as you can easily resize). I think that you can easily solve this by baking certain portions of the switches on the radio, for instance the set on the top left row of the radio. Leave the larger knobs for physical geo and make sure backfaces are deleted where possible. You may have considered this already, and if you have i'm sorry Great work, the only key issue is minimizing your tri count as much as possible and you're pretty much golden.

Hope this helps

-Will

SaferDan

aha after reading Kaburans post i realise i may have been a bit blunt! I apologise!

GarageBay9

Sokay, guys, I really appreciate the feedback.

Yeah, 6k is high for just an environment asset. I intended for this to be a manipulable object, something the player would get up close to and have to flip switches and spin knobs on for gameplay functions. It's still kind of high even for that, though.

vcortis: I actually had the spec much sharper for where the paint had been scratched through all the way to the un-rusted metal, but I pulled it back because it was getting blown out badly. Looks like I made the wrong call. Let me post the texture plates...

EDIT: Yeah, you guys were very right. Without much tweaking I've nuked about 900 tris. Pretty sure I could find more if I spent a little more time.

EDIT 2: THE EDITING: Polycount is down to 2738.

Snader

After having fun for half an hour I came up with this. You need no more then 1000 polies for the thing.

GarageBay9

So you'd bake in the bases and bezels for all the switches, and ditch the holes in the speaker guard?

Spark

GarageBay9, you don't seem to be using any alphas which will easily help you reduce your polycount for this prop. Something I might suggest is have your normal on the object while you reduce, and see where the normal can carry the information. Also I make a referenced object and put a black material on it to look for any silhouette changes while I reduce, if the silhouette is affected I don't do that. But a big thing to look for is piece within the surrounding object, as if you cannot see its form, then you can reduce to a greater degree.

Spark

Progg

+1 to Snader for his model. That's the right idea.

vcortis

Progg wrote: »

+1 to Snader for his model. That's the right idea.

Yeah I concur. Sorry if I came off as harsh earlier too. You really do have a nice start to this, but your optimization needs some work. Something as simple as Snader's model for the low poly should work just fine even as an interactable object.

GarageBay9

Thanks for the input, even the harsh input, guys. I have a lot of ground to cover in a short amount of time (I feed - well, fed my family with this career... been out of work for a while, if I can't get caught up quickly I'm going to have to give up and do something else).

I haven't used alphas much because on my previous projects we always got screamed at by the technical artists anytime somebody suggested using an alpha, they didn't like them for performance reasons. Definitely a habit I need to get back into.

Snader

GarageBay9 wrote: »

So you'd bake in the bases and bezels for all the switches, and ditch the holes in the speaker guard?

Yes, yes, no. I'd use alpha for the speaker cover, and for the wire on the back (so it would just need maybe 6 triangles) I'd probably separate the alpha'd pieces though, so something like a 512² for the main part, and a 128² with alpha. You could also make a slightly bigger alphasheet, and combine other props on it, maybe a ringed booklet with instructions, wires to the mic etcetera.

If you want to model for interaction with the thing, it'd be really cool if you made the mesh/gauze/wire/slits/whatever over the mic with alpha, so you could see through a bit and see the cone etc.

What would also be cool, but it's probably a bit late for that, add some cooling vents on the side or wherever, and have a box behind it with inverted normals and some of the interior.

Like you can see here:



Also, I don't think alpha will cost all that much if you keep it to a simple yes/no check.

GarageBay9

Here is a revised shot of the radio in UDK. I'm having problems getting the specular to behave the way I want - it's either very blown out or very soft. Right now it's soft. I'm probably going to call this good and come back to it down the road. In this version it's 2.7k tris - probably could optimize further, but past this requires more significant geometry and texture changes than I want to keep doing at the moment.

This piece wasn't as impressive as I'd hoped for the job application I did it for, but I'm happy with the overall texture work. I've always thought of my texturing as my weakest skill and this started to change my mind.

Going to go back to blocking the general environment.

GarageBay9

Ok, so a very small update. I hashed out the CONEX container highpoly in max since it's so simple, and I'm baking that right now. The part that's slowing me down is I wanted the player to be able to walk into these containers, and I want to be able to chop holes in the sides to create larger rooms, so I'm having to work that into the mapping...

Here's a quick MR render of the braindead simple high poly. I'll composite dents, dings and other surface details onto the normal with Crazybump instead of burning time in zbrush.

GarageBay9

Ok, UV mapping / texturing theory question.

The player is going to be able to walk into some of these CONEX containers, and I want the texture resolution to hold up at arm's-length distance. I have the exterior faces mapped to a 1024, but I'm wondering how to maintain good res for the interior, because I was noticing parts of the framing for the container are getting really small texel res. A couple of options came to mind, but I'm not sure which one is the 'best'.

- 2048 for the interior (would like to avoid this)
- some kind of layered detail for the interior, possibly vertex colors or polypaint?
- tiled texture for the floor, walls and ceiling, with a 512 specifically for the container's framing (which is the part mainly getting the shaft with texel res)
- make the interior a completely separate object textured at a higher resolution

Thoughts? Not sure how this kind of thing is usually handled, honestly.

felipefrango

I think the interior could mapped on top of the exterior and use a detail normal map for the whole, which is basically a tiling, smaller texture that is overlayed on top of your other textures. You could then bake a lightmap for the shadows using unique UVs for everything.

SHEPEIRO

uniquely tile the entrance then produce a tiled(lengthways) floor ceiling and 1 wall for interior and roof wall for exterior

maybe a 2048tall x 512 map so that most of it can tile horizontally for the walls floor and ceilings and then cram the entrance parts in between

GarageBay9

Felipefrango - so the inside of the walls and ceiling would look the same as the outside, correct? And the detail normal would be mixed into the normal map with something like a Lerp expression?

SHEPEIRO - if I went that way, would I use decals or extra UV channels to put markings / logos on the outside of the container?

Shogun3d

Definitely go with decals

GarageBay9

Gotcha. The extra texture sheets aren't going to be a huge deal? On my last project we were constantly hammered about not using anymore draw calls than absolutely necessary, and my impression was that alphas and extra diffuse maps mean extra draw calls. They kinda took it to an extreme and I'm trying to break myself of a bad habit of not using that stuff, but I want to make sure I'm doing it the right way...

SHEPEIRO

if you do it wisely you can get away with two draw calls one for the main texture and one for the decals.... you could use a second layer in your shader too, which makes the sahder more expensive but saves on draw calls... but i would say that then the entire container becomes more expensive to draw...i would go for the decal option its easier to swap out logos etc you can make very quick variations etc etc

use cuts to get it to tile propperly, but you have to be carefull this way to add a little overspill space for mipping



of course this depends on if this is going to be a single container, or one of many in your scene, if it was say a container city, then i would do things differently, i would put all the different logos on one sheet (all the decals are then one drawcall) all the interiors on one (only need one interior probably) and maybe three different exterior types on one sheet then do another sheet that uses the same spec and normal but has different colours...six colour x decal variations for four draw calls

ps where are the doors and ends for this thing they are usually quite interesting

GarageBay9

Thanks for the paintover, Shepeiro, that's VERY helpful.

I left the doors off because I wanted the layout to be much more open. I figured the inhabitants had taken them off because they didn't use them and they just got in the way. The containers are more for shade and personal privacy than any kind of true enclosure.

I was thinking about having door halves scattered through the scene being used as tables, lean-to's, wall patches, etc.

GarageBay9

Okay, I have one last obstacle with this - how to cut holes in the walls in various places. Rather than building 999 versions with combinations of holes in different positions, I was hoping to use MeshPaint to paint VertexColor masks onto the walls wherever I wanted an opening. Besides being a pain to paint just the wall and not catch parts of the frame, it more or less works. But projectiles still collide with the hole, and I have no idea how to dynamically alter the collision hull so the player can walk through some spots but not others.

Should I build the side walls as separate StaticMesh objects split up into sub panels (about 8 per wall), try to do the walls with brushes, or something else? Being able to re-arrange the configuration of the walls would be a huge time-saver over building all the different wall permutations by hand.



EDIT: the more I think about it, the more I gotta ask. Am I burning too much time on this? Should I just make it 'good enough' and get my butt in gear with the rest of the scene?

Knowing when to put something down and move on is something I'm always struggling with and I can just never tell when I'm overdoing something or when I haven't pushed it far enough.