Environments questions for nextgen artist
I was just wondering if someone out their in the industry could tell me a couple of things about current next gen environments and props. I'm currently looking for work(mcfarlanemodels.com!...lol, cheap plug) and working on putting together an environment portfolio, but don't know the current polylimits and such. If anyone could tell me
Average polycount for game level?
texture sizes and polycounts for a wall, building, or props?
Is everything normal mapped in nextgen games, or is it limited to mainly characters and certain props?
last one-Should I paint the chrome highlights on objects since many games have dynamic lighting, or leave reflections out? (Example, I'm making a low poly locker prop, without painting the relections and highlights on the silver handle, it dosn't look "silver", do I leave this to the shader?)
Hopefully these questions arn't too stupid, and someone can answer them. Thanks to all who reply
-justin
Average polycount for game level?
texture sizes and polycounts for a wall, building, or props?
Is everything normal mapped in nextgen games, or is it limited to mainly characters and certain props?
last one-Should I paint the chrome highlights on objects since many games have dynamic lighting, or leave reflections out? (Example, I'm making a low poly locker prop, without painting the relections and highlights on the silver handle, it dosn't look "silver", do I leave this to the shader?)
Hopefully these questions arn't too stupid, and someone can answer them. Thanks to all who reply
-justin
Replies
I don't paint much in photoshop anymore. If you want chrome - use a reflection map - just like you would in pre-rendered texturing.
Most of the small to medium objects are about 300-1000 polys. Each window in a building might be 800 polys for example. I have no idea how many polys there are in total.
First and foremost, focus on what you want to do. Most studios have prop artists and level artists / environment artists. If you want to do props, focus on those and make them the best they can be. Environments are much harder, as one piece can take a whole lot longer.
General rules - yes its important that you can stay within poly limits and texture sizes - don't go crazy but show them what you can do. You never know when an AD may ask you to beef something up for a cinematic.
Lastly, shadows and highlights should be separate from your diffuse map. Shadowing should not be baked in but rather an end result from your normal map and a light w/ a shadow for props. Highlights from your spec map. Reflections from a shader.
Sorry for the long post! Oh, and sprunghunt, don't you think 800 polys for a windows is a bit over kill?
You do what gives the best result in the shortest amount of time. Learn how to use the PS filter to your advantage. Its a great tool and espceially in combination with generating normals from high res.
About polycounts and texture resolution its very hard to give an exact answer since it depends on what engine and what game you are working on. In my experience what you can get away with now is a higher polycount. Texture resolution is not so much higher than it used to be. Remember for every flat diffuse you used to have now you got atleast a diffuse, normal and a specmap to boost. Again this depends on the game engine.
I think you should look at a game you like and try to mimic their looks. Keep your polycount and texture sizes reasonable and you will be fine. When hiring for jr. positions most game companies want to see you can get the job done. Specs for their engine you will get on site.
About the chrome you definitly want to leave them out. Leave that to your specularmap.
Texture sizes? It depends.
And those are the two most honest and realistic answers.
As Rick so eloquently put it, there is a huge amount of variation on polycount and texture limitations, depending on the hardware that you are developing for, and the type of game that it is. For games that I am working on, a large prop such as a small exterior building model would be under 2000 polygons, and would have diffuse, specular (possibly spec power as well, depending), normal, and lightmap/ambient occlusion textures. I would author at 2048, and ship 1024 tgas. I'm sure some of these maps may be removed during optimization, but we always supply them in case they are needed. Note that lately we have been using mostly single texture sheets, rather than multiple tiling textures for most props and buildings.
Shaders and specular/reflection maps will handle shine and reflections, so avoid too much painting of highlights. Also, don't paint in too much shadowing, as an ambient occlusion lightmap bake will handle most of this as well.
Sorry for the long post! Oh, and sprunghunt, don't you think 800 polys for a windows is a bit over kill?
[/ QUOTE ]
It says in the title that it's Next-gen
I assumed this meant at least an XBOX-360. For that platform an 800 poly window is not overkill for a first, or third, person style game. It would be overkill for a top-down strategy game.
Oh and this also assumes that your window is one that the player can get close to. Buildings that are in the background just wouldn't have separate windows.
I'd suggest checking out the Gears source if you can - it's very instructional.
And I would contest that for realism every surface needs a normalmap otherwise the lighting doesn't look consistent. However the proof is in the pudding as they say.
[ QUOTE ]
Sorry for the long post! Oh, and sprunghunt, don't you think 800 polys for a windows is a bit over kill?
[/ QUOTE ]
It says in the title that it's Next-gen
I assumed this meant at least an XBOX-360. For that platform an 800 poly window is not overkill for a first, or third, person style game. It would be overkill for a top-down strategy game.
Oh and this also assumes that your window is one that the player can get close to. Buildings that are in the background just wouldn't have separate windows.
I'd suggest checking out the Gears source if you can - it's very instructional.
And I would contest that for realism every surface needs a normalmap otherwise the lighting doesn't look consistent. However the proof is in the pudding as they say.
[/ QUOTE ]
what do you mean exactly when you say window? if you mean the window glass, and frame, then i still agree 800 may be a bit overkill.... however yet again thsi depends on many other factors.... how big is the window? how many other polys are in the environemtn already? how many moveable/dynamic objects will have the possibility of being onscreen at once (characters, vehicles...etc)? and what hardware, however you've established the 360, but even next-gen games still try to support lower-end hardware if it is on the pc platform. these are the questions i have to ask myself when doing environments for my current project, i have the possibiliy of 16 humans being on screen at once and more polys depending on thier equipped weapons, we sre trying to use a lot of shaders and have the game look very good yet support the m in hardware of 5700...etc. for my game 800 polys for a windows would be overkill.... but, for say, a first person shooter, it might not be depending on the other factors.
i can think of no reason that a window/windowframe, even with modelled breakable glass, should be as high as 800 polys.
I think there's a lot of variation in what people are thinking of as "a window".
...and you dont need to generate all your flat textures from hires meshes, that's just a waste of time.
As for polycounts in next gen games, just grab one you think looks good and is kinda what you'd like to work on, play through it or look for screenshots, dissect it and go from there.
I think some people have exagerated views of next gen stuff.
Just because you technically can detail something all to hell doesn't necessarily mean you should. Now if you're doing dozens of uniquely-styled buildings, what is more important? Getting them all done, or finishing only one with kick-ass windows with details that most players will not notice? Sometimes just a quickie grayscale bump to Nvidia-filtered normal map will provide the enough of an impression of detail to be convincing at a glance. And it'll take a fraction of the time and polygon count than would modeling, UVing, and texturing every little detail.
With the more ability to detail things, it's easy to lose sight of what the real priorities are. The biggest real priority is to get everything done.
http://www.rsart.co.uk/2006/11/20/how-many-polygons-in-a-piece-of-string/
Giving people a few tips and means on how to find the information they seek, would be nice.
You do not have to generate all your normal maps from a high poly model. I want to be clear on this. Certain factors come into play here; for one is the length your project. Can you afford the time to highpoly model everything down to a loaf of bread? Not if its a 1-2 year development time. That would be a waste of time anyway. There are certain things that you want to choose (round surfaces, organic/faceted level of details that cannot be done as a hand-painted normal map). Use the Nvidia Filter or CrazyBump to you're advantage. Do tests for yourself; select an object that isn't insanly detailed and do a highpoly, texture it, etc. Then do the other one where you just made a lowpoly, then textured and did a hand painted normal map. You'd be amazed at the difference in how long it took and the level of difference it looked. You begin then to really get a sense of what does and doesn't need a highpoly source.
For textures on large scale architecture; try to make textures that can be tiled for better resolution and re-use in other areas. You don't need to do UV layouts for everything; just projection map it then tile. Use various decals to help break up the tiling if it looks too repeated. All this stuff alows tons of resolution with a smaller library of textures.
I'd also start thinking of terms of modular pieces. Figure out how you can model a roman cathedral with 30 or less peices of geometry that you can rotate/scale/stack. Shit like that. The re-use/instancing of geometry and textures helps performance in addition to memory a lot.
Employers are going to want to see quality, but they are equaly interested in how fast and efficient you can do it as well.
indeed... i dunno where that figure is coming from but either you're modelling every pane of a stained-glass window, or you're making numbers up off the top of your head.
[/ QUOTE ]
I got the figure by bringing up statistics on the Xbox 360 dev kit which was running our latest game.
That's why I think it's a reasonable figure. Because that's what we can sucessfully use in our game.
I just thought that since an aspiring artist doesn't have access to dev kits that maybe I'd give a hard example that might be useful rather than the usual vauge advice of "it depends". Anyone else want to give some figures?
"Its rarely *just* about polygon budgets either, its about texture budgets, bone numbers (for animations), material types (which governs how many draw calls there are for the model) and re-usability of that model / material within the scene (usually zero, but it can be a mitigating factor.
A short answer is as few as possible without sacrificing the minimum quality bar thats expected for the game, but even thats pretty long.
Ultimately, its *very* easy to do great looking high-polycount models, and *very* hard to do great looking low-polycount models and the success or failure of adhering to poly limits set by your lead art (or code in some cases) can seriously affect the quality of the art you produce and the performance of the game as a whole. "
Alex