Texture modding

[csgators]

So I extracted the urbanposters01.dds and edited it. When I save it says that this image appears to be a volume map but not all layers are the same size. This is true, not even sure what the built in layers do but I left them in place. When I save anyway the posters either don't show up at all or appear and disappear as I move around. Any Suggestions? Can I delete the 10 or so built in layers? Do I need to do something before I save? :wallbash:

[Ghogiel]

I'm guessing you are using the gimp? The file was likely orginally saved as a dxt1 1bit alpha- it says so in one of my programs, You can trust that if you want.

I guess you mean the built in layers would be the alpha channel. I never used the gimp, but from what I have heard, it doen't handle alpha channels like mainstream bitmap editors, photoshop/paintshop/etc. So perhaps its that. to answer your question- do you need them? if you want the transparent parts of texture- yes. I would look at how to deal with alpha channels with your program.

 

The other thing you mention- "appear and disappear as I move around" could that be archive invalidation?

[csgators]

Yes I am using GIMP with a DDS plug-in. I just removed all but the top layer and saved with no problem and the posters seem to show up fine now. Not sure why the original had 9 seemingly useless layers.

 

dxt1 1bit alpha

 

Whoosh, right over my head. :biggrin:

 

if you want the transparent parts of texture- yes. I would look at how to deal with alpha channels with your program.

 

The alpha is the transparency right? The rips in the posters seem to work (got a lot of work to make them look good though).

 

:thanks: for the help, I need to read up on this stuff I guess.

[csgators]

I think I see what's happening, when I open the file it asks if it should create mini-maps. If I say yes it adds 9 useless layers of different sizes.

[low8all]

I think I see what's happening, when I open the file it asks if it should create mini-maps. If I say yes it adds 9 useless layers of different sizes.

 

Right, mipmaps are LOD image maps that are used based on the players distance from a given texture. The further away the player is, the lower the LOD (the smaller the detail is needed)

 

On the topic of alphas. The posters have an explicit alpha channel. You'll see it when you view the channels, labeled as alpha. It'll be white where the texture will show and black where it wont. Since the posters have torn edges and holes. You can change that just by painting on the alpha layer with black and white - black for see-through, white for opaque - and you can feather it of course for a smoother transition between the two (but the quality of that transition will be based primarily on your export settings and secondarily, on the users settings - so high quality setting will notice a much smoother/better transition between what's in alpha and what is not).

 

When you save, to get a smooth transition, save as DXT3 ARGB 8bpp | explicit alpha. This will retain the feather between alpha and opaque (if there is one) and use the alpha channel explicitly - as it exactly is. Saving as 1bit alpha will create the alpha based on the 1bit pixel point on the alpha channel. It will either be alpha or it wont be. Which will jag the edges of any transition between 100% alpha and 0% alpha.

 

Keeping in mind, of course, that the explicit version will be slightly bigger in file size but for posters, it's negligible.

[csgators]

Thanks, that is very helpful. But does that mean that I need (or should have) those min maps? If so why can't I save it with the mini-maps in place?

[low8all]

Thanks, that is very helpful. But does that mean that I need (or should have) those min maps? If so why can't I save it with the mini-maps in place?

 

When you import the .dds file? No. You can set the import to no mipmaps. Mipmaps are made on export by the dds exporter (if you have the option selected - which you should, generally speaking). If a dds image has no mipmaps, it will just be rendering the default version.

[csgators]

I get it now (the mini maps at least) and I found the save format you recommend. Thanks for the help, I am very new at image editing but I'm learning.

[low8all]

I get it now (the mini maps at least) and I found the save format you recommend. Thanks for the help, I am very new at image editing but I'm learning.

 

No sweat csgators.

 

To be honest, image editing, from a modding standpoint, is probably the easiest form of modding. Scripting and modeling/animating have huge learning curves while texture modding is singularly image-centric. It will either look good or it wont look good. It will either show up or it won't show up. It's nowhere near as complex and multi-layered as scripting or 3d but can, in most cases, have as big an impact and fundamentally change the game-play experience. (when done well and to the extreme)

[Ghogiel]

When you save, to get a smooth transition, save as DXT3 ARGB 8bpp | explicit alpha. This will retain the feather between alpha and opaque (if there is one) and use the alpha channel explicitly - as it exactly is.

nope. dxt3 only keeps 16 values of alpha. http://1.bp.blogspot.com/_ePzz13VNUS0/SwFyBg8fzVI/AAAAAAAABAQ/9hCCISZDYm8/s1600/grayscale.jpg

 

if you have any value in the alpha besides one of those 16, it'll be resampled to the nearest, creating a steeped effect. used for sharp alpha transitions.

 

if you want to retain you're alpha without reducing it to the 16value palette, use dxt5. used for smooth gradient alpha transitions.

 

the file sizes of dxt3&5 are exactly double that of dxt1(with or without alpha). 8bpp vs 4bpp. 4:1 and 8:1 compression ratio respectively

 

tbh- you never need to use dxt3, If I want to reduce my alpha images to a limited palette, I can very easily do it with 1 simple click, and still use dxt5.

 

in this circumstance, i would use dxt1 1 bit alpha- for one it'll save mem. and 2, at no point would a ripped poster have semi transparency around the rips. it'll either be on or off. if it was an eithe/or issue- using dxt1 over dxt5 would tempt me to just double the file resolution, and then the alpha map wouldn't be as pixelated, and would allow me to really get pumped on the rgb channels :cool:. between the 2- I bet that dxt1 set up would yield the best result.

also don't neglect the nialphaproperty flags, and its threshold. it's always worth a play with.