Game texture questions

[KazuyaMochu]

ok... in sequence:

1: When you paint a texture for a game, do you usually paint it at a high res and then shrink it down? or do you find that you can get more detail if you paint it at the res you will render at?


there is no agreeing for this. alot of ppl say that you should allways do a double size texture (just in case...) and then shrink it down. In a certain way, this is good, becaus working in a bigger size, you have room for lots of mistakes and so.
working on 1:1 you ar acctually working on the final texture. I meen, you will preview a little better how the texture is going to com out. when you shrink a texture, you lose detail. that you will have to put in the smaller version again. and don't go for "filters>sharpen>sharpen more" that usually wrecks it.

a good way is to get things layed out in double size. text it on the model, and when you start adding detail, you should shrink it.
some times a glow is only a small white pixel!

oh, and allways save frequently

2:is there such a thing as LOD textures?

I think your talking about mip mapping.
I'm not sure, but I thin mipmapping generats several resolution textures for the same model, and uses them on the engine, acording to the distance.
it also inproves de aliaising and overall look.

3:is it more efficient for the renderer to have one big texture, reused for several objects than to have seperate, low res textures for each object?

I think it depends on the grafic engine. for example the unreal and quake engines, have several little texture for every thing. But they all belong in the same file.

Gta 3 has sereval little imagens, belonging to the same file, in the big file (that has several diferent files inside)

(man that must be confusing)

But from my experience, you shoul try to put them together as much as you can, in a reasonable way. for example: a house. it isn't very good to have a texture for the ceeling. another for the window. one for the floor. another one for the door... you get it, right?

but in terms of level design, and funcionality, some textures shoul be alone in there file. for example, floors, or wall textures, because they tile alot. and not being separate, the modeler would have to put aditional poligons to be able to create a sort of fake tilling.

I think you should get the not tiling texture together, and leave the important one separate. even in therms of organazation, it works better this way.

4:is it *always* necessary to have a square texture, or does the "powers of two" rule work in ratios? like are 2 512 x 256 texture any less efficient then a 512 x 512?

In most cases yes. you can also have 512x256 ou 128x1024 texture sizes. but its better for memoru alocation, and for example photoshop even has filters working for thoes sizes. like clouds. if you put clouds in a power of two image size, they will tile perfectlly.

if you engine alow other sizes... they will probably go to "dividable by two" sizes. But I thinks its better to keep within thoes sizes. that way, you make them fit any game engine.

(man that was long!)


hope this help. also hope not to have confused you or something...

Kazuya Mochu
:thumbsup:

[Jboy]

wow, thanks! that was exactly the answer I wanted. Very informative.

this internet thing rules :buttrock:

[kwshipman]

Originally posted by Jboy
this internet thing rules :buttrock:

now if it will only catch on. Some day my friends, some day.

[bugo]

Obrigado pelo tutorial ae Kazuya, você escreveu "ou" ao invés de "or", foi muito engraçado, heheehehheehehe! :applause:

Thx for the tutorial Kazuya, you wrote "ou" in turn of "or", it was funny, heheehehheehehe! "ou" is "or" in portuguese language... :applause:

[BiTMAP]

I would still support working at double size becuase in photoshop when you size down the details are still kept in a lower res form. You draw a small clip on a shirt, and size it down that clips pixels are blurred and loose some of hte colour value but at a distance you are tricked into seeing the clip still.

something i said when i was working on a pixel art peice i have.

You can't make everything, so make it look like everything.

[KazuyaMochu]

bitmap, I actually would have to agree with you. But my answers were basiclly the general discussion resault.

I still belive, that after you reduce it, you should double check the details. and you will probably redo some.

but yes, photoshop as some great resizing algoriths.


like this imagens I made, from hirez drwaings or renders, into icons. phoshop kept most stuf in. I just had to adjust de contrast.

[BiTMAP]

mm those are nice :D lol.

actualy yes, often you do have some adjustments after you resize, I actualy sometimes paint sections over again but for the most part the size down is good enough.

[CADster]

i perfer to do textures at res.... but doing them double size and then rescaling enables you to add more detail (maybe for the render shots, cover art, etc.) so doing them large has its place.

a new image format (DSS, or DDS, i forget) is creeping into a ton of games. it has mip imbedded and its very cool.

[KazuyaMochu]

Yeah... DDS

it's pretty cool for PC and XBox, but PS2 does not support them. Png is the optimal format for PS2 games I think.

[AC_Shrike]

JBoy,
The replies here are all pretty valid. However, I've found, personally, that in some cases working at double the final size and then sizing down can cause much of that detail you work so hard to create, to just close out and get fuzzy. For example, a single pixel line, when sized down from a larger image, can and most likely will lose its crisp edge. I do create at larger than working size often, though. For instance, on my current project, my game textures are created at 768 x 768, then sized down to 512 x 512 (our game engine is quite robust and processes the textures quite efficiently). Working at final res - creating your textures at the size for which they'll be used - is often the best method, but it can sometimes be quite painstaking to get the detail you need, resulting in pixel by pixel editing. Experience will help you make these decisions and every artist has his or her own way of creating, but everyone here has given you good solid standard guidelines.

Textures can, of course, be different dimensions, such as 64 x 512, 128, 256, etc. but yes, sticking to a 2-power ratio is optimal.

Finally, try to use the least amount of textures as you possibly can. I'm learning the hard way than most if not all game engines are less likely to choke, the smaller the texture memory allocation that has to be used. Efficient uv mapping is key here. Good luck in your endeavors, my friend.

[Neil]

I work at the file size i plan on using.
I don't see how you can say you can add 'more' detail if you work on larger size.
Either way the pixels are still gonna be filled. You just have to zoom in and work on a smaller scale if you work on a smaller canvas.

[Kanada]

Try creating your textures at 256.256 and then reducing the colour depth to 8bit, then reduce the size. As the texture will be filtered ingame you will mostly get a better result than if you reduce the size at 32bit.

Mip mapping is not just done for visual reasons as if a texture that is 128x128 is displayed on screen in only a 8X8pixel area then you will get slow down. It would have been more efficient for the texture to be uploaded at a smaller res. :hmm:

Also there will be an optimal size to upload a texture into Vram depending on which console you are working on. So lots of small textures would be bad, rather combine all those small textures into a few larger textures.


:surprised

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.