Page 5 of 20

Posted: Fri Aug 11, 2006 2:30 pm
by Dogbones
How do these new textures look when sampled back to 256^2 or 512^2, which is likely what will be happing for a good while.

I guess my point is, we don't want them looking great at 2048^2 or 1024^2 only to have them look worse than what we have now until systems that can handle the higher textures are the norm.

It would make for a lot of wasted effort if a 2048^2 texture (that took days to make) looked bad at 512^2, when a 512^2 version could have been made from the original 256^2 in a fraction of the time.

Posted: Fri Aug 11, 2006 3:17 pm
by Zapper
Dogbones wrote:QUOTE (Dogbones @ Aug 11 2006, 04:30 PM) How do these new textures look when sampled back to 256^2 or 512^2, which is likely what will be happing for a good while.
I guess my point is, we don't want them looking great at 2048^2 or 1024^2 only to have them look worse than what we have now until systems that can handle the higher textures are the norm.

It would make for a lot of wasted effort if a 2048^2 texture (that took days to make) looked bad at 512^2, when a 512^2 version could have been made from the original 256^2 in a fraction of the time.
Yeah i saw that comming and i need to ask YP how he wish to handle the diff texture size's
If the engine downsize the texture from the 2048 or we are going to make a lot of textures with diff surname texx_512 etc.
We can downsize the textures or simply resample the tex from the begining and work on each set of tex as a project on its own.
But that is a huge workload and i would not do this cause we dont have the man hours to cope with this task.
So in order to evade a huge work load we are still needing a High rez Tex as a master textures.

If the engine is downsizing the 2048 tex to the target size, we would get a problem cause we would still load the full file size into the cache on the GFX card.
The outcome depend on how the engine will handle the diff resolution's.


This is a test we are gonna do on the astroids, so we have knowledge on how it will work out.

Tell me how the engine handle the textures and ill work out a solution.

Zap

Posted: Fri Aug 11, 2006 9:17 pm
by Dogbones
I'll let Y_P repond regarding how the downsampling it done.
AFAIK it is done at load time when the texture is read off disk and loaded into texture memory, so it is not being downsampled by the card. This is also done automatically, so we won't need a bunch of different files.

My concern was that a 256^2 texture nicely redrawn at 2048^2 could look crappy when resampled to 512^2 while one redrawn at 512^2 would look better.

It boils down the how much of the fine detail is critical to the scene. If you make nice fine hull edges, when downsampled that could look pretty bad, with gaps, etc.

I work with medical images and scanning something at 2024^2 and then down sampling to 1024^2 is not the same as scanning it at 1024^2 in the first place. The downsampled image is almost always worse. Now we are not talking about medical images. I just wanted to raise the point.

Posted: Fri Aug 11, 2006 9:40 pm
by Adaven
Well as far as downsampling concerns go, I don't think we have to worry too much at the moment. The first 2 items on most people's lists are Environments and Asteroids, which are very, very forgiving objects to texture. A lump either here or there makes little difference. It'll still look like a piece of rock or cloud of gas.


For stations and ships we'll probably need to do some testing to make sure everything looks good. I'm sure that a native 512^2 could be made to look better than a downsampled one, but there might not be much difference when it is all said and done. We'll just have to play around with them once folks get some more textures cranked out.

Posted: Sat Aug 12, 2006 1:10 am
by Adaven
Here's bgrnd03, this and bgrnd05 are all the generic asteroids I think, just tech and he3 rocks remain.

http://i42.photobucket.com/albums/e313/Ada...grnd03-big2.jpg

Again, some source material courtesy of NASA/JPL-Caltech.
Now that I have a better idea what to do, this one was much easier and has more detail in the right places that the other one I did. I really like making rocks.

I've still got a seam issue I need to take care of, but it shouldn't be too hard.

I'd start sending the files to whoever wants them, but I'm on dialup until the 21st and each one tops out at 8.2 MB.

Posted: Sat Aug 12, 2006 2:12 pm
by Weedman
I'm going to need a new video card.

That is one beautiful ice rock adaven.

Make one with some debris from my kills embedded in it, and model a little flag, it reads, "Weed was here".

Seriously..

I have a question. Is 2048 what you're going with for the high res?

A number of the textures for starwars will fit on a 1024, and yet there are many that will not, I have had to put them on 1280, but I would have to put them on a 2048? That's alot of wasted space. And this high res is optional? My only concern is putting alot of time into the high res textures, and when somebody wants to use 512 or whatever, it looks like @#(! using the high res coordinates, technically it shouldn't change, but it does, the pixels get moved around and what looks great and correct in high res, gets fudged around in low res.

Posted: Sat Aug 12, 2006 6:30 pm
by Orion
Maybe make the rock less.. brown. That would help with the realism.

Posted: Sat Aug 12, 2006 8:35 pm
by Adaven
Orion wrote:QUOTE (Orion @ Aug 12 2006, 01:30 PM) Maybe make the rock less.. brown. That would help with the realism.
You are probably right, but for the moment I'm just been sticking to the colors on the original texture. That particular rock actually has some grey splotches in the original. I tried to match them up with larger craters to suggest that the color distinctions were the result of newer material being uncovered by impacts. Regardless, the cratering I added did darken the overall texture, possibly muting some of the color differences, I'll try to brighten them up in the next go-around. Sector lighting may also be a culprit, I'll admit that wasn't the best composed screenshot. This may be one of those deals we'll have to let people test out in game to know for sure.

But back to the issue of color in general. If the generic rocks being predominately brown isn't realistic enough, what do you think of the tech rocks? Are Silicons too green, and would a high concentration of Uranium actually be orange? All the real images of asteroids I've come across are predominately grey, with any significant coloration being the result of enhancement.

Anyways, thanks for the input, I'll try to make some color adjustments after I get the basic patterns done for all the rocks (4-6 I think).

Posted: Sun Aug 13, 2006 1:05 am
by jess_i_74
That looks fricking awesome. Nice work.

2048 does seem a little extreme . . .

Posted: Sun Aug 13, 2006 6:26 am
by Da_Muck
I think part of the problem is the in game lighting Adav... The current common lights are what turn the greys more brownish, maybe? The actual detail in the texture looks great though.