A mana well
I'm struggling with textures as these textures were from cg textures.com
I think i'm rushing my work.I read what Vig said about 4:1 texture ratios and It's something i did not know what so ever.I really appreciate tips like that.So i'm posting this here in hopes of more advice.The texture I used was a 1024x1024 texture.It's for a small asset in a tower defense game.I still can't wrap my head around the 4:1 thing.
If my model's height is 256 then my model should be placed on a 1024 texture?If it if my model's height is 512 then everything should fit on a 2048 texture?
I think i'm rushing my work.I read what Vig said about 4:1 texture ratios and It's something i did not know what so ever.I really appreciate tips like that.So i'm posting this here in hopes of more advice.The texture I used was a 1024x1024 texture.It's for a small asset in a tower defense game.I still can't wrap my head around the 4:1 thing.
If my model's height is 256 then my model should be placed on a 1024 texture?If it if my model's height is 512 then everything should fit on a 2048 texture?
Replies
1) You really should stack UV's all those wood part can share same space.
2) Stone part also can share same space
3) I don't think it's good idea (depends on where you will use asset), but you can also use sub-materials for biggest part (single tileable stone texture)
4) You should also use symmetry and make only quater of those fountain, well or whatever ;p.
Essentialy making environmental assets is not aboutcreating one huge mesh from single primitive, but making as little as possible geometry and reusing it whatever you can. Also about textures you should share textures with as many assests possible, which means you don't draw texture into your unwrap you unwrap model to match your textures.
This of course depends on how much "uniqeness" you model must have. But you still should think about how to split your mesh into smaller parts, and share single UV space. As example. You don't need to make entire Cylind it will be enough to make 1/2 1/4 or even 1/8 of it.
So if i was to model a car I wouldn't texture it based on my unwrap?
I have to remember to use symmetry and mirror my uvs. One of these days I'll get it.It's very frustrating.
http://www.roryjones.net/gallery/hoverboard/#hoverzflats.jpg
Hope this helps and keep at it!
No. Car is whole another thing. But for environmental you will end up doing it this way, because you need to as few as textures possible, that are also pretty low res.
also, there's no reason to do 4:1 texture rez on an asset if that's the viewing angle you see it at, that could use a 512x512, 256x256 or even smaller.
(this is quick and sloppy cuz I only had a few minutes to throw it together, don't just throw a photo on the model and call it done, paint that shit!)
Did you hand paint that texture or overlay an image?Someone told me you could get fired for using textures from an image and you should only use them as references and manually paint everything.
Why did you paint in the highlight I thought you were suppose to let the program do that and leave it out of the diffuse?(looks great though)
How would the other assets fit into your texture space? I would have to have seperate textures or multi sub texture it.
It's not as clear cut as height X 4 = texture size. The examples I was talking about in the other thread where about a box which translates pretty well because it fills its volume. But your shape doesn't fill the entire volume, you have a few thin pieces sticking out of a low base shape, it's a pretty empty cylinder volume wise.
You have to think of it a bit more like you where carpeting or painting an oddly styled house. You have to think total surface area of the pieces, not volume. If you unwrapped those pieces in the game, laid them out on the floor what area would they take up? Then you can times that by 4. If you're using a 4:1 ratio for your game, the game might use something else entirely, it depends on the game.
Easier said then done I know...
You can make things easier by eyeballing it. It doesn't need to be pixel perfect and 100% mathematically correct, if you eyeball it close enough.
But how do you eyeball it?
You unwrap your pieces so that if you apply a checker pattern the squares will be roughly the same size. The default procedural pattern works ok for this.
[pause]
These scripts won't figure out the texture size but they will help you make the size of each piece consistent with respect to the others
Once its all unwrapped and the pieces are scaled appropriately (in relation to each other), then its just a matter of figuring out what size to make the texture.
You can apply a custom checker pattern with a known grid size such as 4px squares on a 512x512. I suggest using an actual image instead of the procedural checker pattern that most 3D apps use. After you do it a few hundred times you get a good feel for it.. Its probably a good idea to create custom grids for most of the standard sizes 64, 128, 256, 512 ect.
Then guess on the right size based on what you think the surface area will be in game and what power of 2 size would fit best. Apply that checker to your object.
Create a cube set to 1x1 game unit. If they roughly match then you're in good shape.
If they don't you need to adjust the size based on the power of 2 chart or think about overlapping and scaling your pieces to get each UV piece to take up more pixels on the same texture sheet.
If I had to guess For your well I suggest using a much smaller texture. Probably 256x256 or maybe even smaller by stacking.
You'll need to rework the geometry on the stone to be taller and more well like.
Then maybe use the bottom 2/3rd for the stone and vegetation, and the top 1/3rd for the posts and wooden parts.
it's really how big the asset will be on screen, it's the same as "how many polygons should this be" the answer "it depends"
I just used a cgtextures real fast because I just wanted to whip something up in 2 minutes. It's fine to use cgtextures, it all depends on the style. Sure, you can get in trouble if you just go and photorip something from google image search but places like Cgtextures exist for this purpose.
because I was just whipping something together fast, I just shit something out. You posted a hand painted asset on the end so I figured that's what you were shooting for. Your asset right now looks like straight up photo sourced.
I'm starting to wonder if I'm wasting my time pursuing this art, you just sh**ed that out?Your work makes mine look worse than sh**.Like yours just took one.This is sickening.
lol, what ? The guy who said that should be fired for saying that, or is that something you thought you knew without anyone confirming it ?
whatever works as long as it looks good, and doesn't take you a bloody month to texture a damn 1024, period.
Justin's been cranking this stuff out for a while now. At some point it will be muscle memory and you'll wake up to finished assets sitting on your desktop like he does. You'll scratch your head and think "elves?" "how much did I drink last night?" Then you'll realize you did it in your sleep.
Quality over speed. Speed comes with doing quality over and over again.
There is one thing in life that will knock you down for good. If you're left wondering was that it? That was not it.
You'll get there, but not if you give up.
That prop looks like something that would be in a 'hand painted' style. So hand painting it would probably look better than sourcing textures.
Usually some highlights/shadows(ambient occlusion) are good on most models, then again depends on the style you are going for. handpainted games like Torchlight are probably more inclined to have painted in highlights than realistic games like Halflife and are less reliant on in game lighting to do the job.
Finish this up, make it look good, learn to make a clean 'spec sheet'. Pic of rendered model, wireframe, some text details (author, tri count, tex size..)
Simple is best.
Once you get several pieces you're probably ready to start showing them to possible employers.
Apprenticeship to me is just a nasty word. Who can afford to work free?
And the TF2/Polycount pack shows the respect that polycount has in the gaming community IMO.
You deffinatley can't go wrong by posting stuff here and listening to what 'most' people say. Of course it's a forum and you can always recieve bad advice, but I think on the whole around here 99% of it is good and if someone gives bad advice someone else will point it out quickly.
But for someone to tell you exactly what to create though, might be a little much without details.
What game
interested in charcter design/props or level design
what style, realism, painterly
Those details go along way. Some companies do every type of game possible (Valve: Halflife-realistic; Team Fortress2 - painterly, cartoony). And that's more important than 'look, this guy does great chairs'. They want you to be able to do a wide variety.
So probably a good start is going with what you like/what you are good with.
If you like Torchlight you can easily place props in the game engine to see how they look. Same with Halflife, Unreal,...the list is very long. I'd say 50% of games anymore have the possibility of modding.
So if you like one specifically make content for it as a showcase piece, also shows you can tackle the editor which is probably important to most studios as well.
I can't give you work related experience on what studios want but I do understand what they want.
They want you to:
know how to use at least one 3d program well. (specifically things that reference the job you want)
Know how to use photo editing programs
be able to stick to/create concept art
be able to communicate
using an editor is a plus.
of course the more you kow, the better you are...
But alot of studios will hire if they like your style/it meshes well with theirs, or at least the game they want you to work on...
So you might prefer to stick to painterly style, and might not have as many opportunities as someone who can do it all. But if you are really good at painterly you might be hired for that reason along.
I think most people can do alot, but focus on one area, that way they get better at that one area and the studio can depend on that.
At times I wish I could bookmarks posts
http://www.polycount.com/forum/showthread.php?t=46693&highlight=DDS
I think that's what snowfly is talking about? Pretty good method since most people are in and out of the engine/game while they create its a good step to keep in mind. I like it!
But it depends on how easy it is to drop an asset in engine then view it in game with the correct texture. In source that requires 2 text files to be written and the model and textures to be compiled.
Sometimes its easier to drop a grid on it and checking its size to a cube.
After a while you get a feel for the right size and I imagine you don't do too much of either, your guesses just become so close no one cares?
Thanks for the link, instead of writing giant walls of text I'll just point people there from now on
Create a 2048 pic in photoshop, save is as dds with generate mipmaps on. Re-open it in photoshop with load mipmaps on. Then edit the pic the way you want and save as dds with use existing mip maps.
/Just seen Eric made one in the linked post.
And FWIW, here are the auto-generated checker images from Blender, plus a mipmap tester DDS.
http://dl.dropbox.com/u/1751508/TestUV/Archive.zip
Any updates on the mana well?
I also thought my 1024 was in fact 1024,that's jacked up.So much for making my own.Do you have one with all of the dimensions stacked?In the fashion i was going for?
You won't want to apply the dds texture to your object when you're editing it.
As you move closer and farther to the object in the game, its going to pick higher and lower textures to display. The one it displays most often is the one you use.
I was playing around with the DDS method in UDK and it works pretty well, I like it. But it was a bit of a pain to set up the first time around. Next time around it will go faster because I'll already have a test scene step up, and the texture build in the material editor. I'll probably do this for each of the engines I work with, Source, UDK and Unity.
The other method (custom checkers) requires some prep too by requiring you to sync your 3d app unit scale to the game which in most cases the default max/maya unit is the same as the game, already synced... and you need to create the materials for testing.
Is there a good site to find a mod group, when your just starting out?I found moddb.
Unreal, Cry Engine, Oblivion, and Fallout groups if possible.Seems like alot of groups are already established.(as in everyone is experienced)
There are a lot of good groups out there that are looking for volunteer help, but there are also a lot of awful ones too. If you get into a bad one, you'll find that they'd be asking you for an inordinate amount of work to be done, they'd have a lot of long drawn out concepts that aren't coherent. There won't be a lot of other people working on it to help manage the workload as well. It's not a great situation to be in, so really pick carefully.
I'd even suggest that you post a couple websites of projects you'd like to apply to, and let us take a look at it to see if it's a worthwhile place to spend your time.
I haven't received a response from anyone.but it'd be nice to join up with some people with know how and work on some in engine stuff.
Especially if I'm capable of picking up on Fallout's Engine and Oblivion's engine as I'm in Maryland.Right now all i know is Unreal.
We're always willing to except models for The Dark Mod (see sig), Doom3 engine, heavily modified GTK Radiant editor (very similar to Hammer- l4d, hl2, tf2). And I'm willing to help you learn how to get stuff in game (model formats, shaders, entities, etc...)
But it's medieval/steampunk/FPSneaker... Might not be your cup 'o tea. We're not looking for team members anymore as we've already released beta, but we are updating fairly regularly, adding content, crushing bugs, etc...
Nothing too serious as stated above, we can help you lern the stuff, if you can make good stuff to contribute we'll add it (it's all open source so your stuff would go under that clause, anyone can use it)
There will be enough people telling you what to make and how fast you should make it later on. Enjoy learning and being your own creative boss while you still can =P