- 492
- France
@Langheck_917Hello,
I'm not an expert skinner and am only talking from my own experiments point of view.
- Using mipmaps increases skin size by 25%. I have no definitive idea about having them stored in the dss or not. I've read somewhere you can safely avoid mipmaps, Assetto will then generate them at loading time (increasing a bit loading time) . Lastest RSS releases don't use mipmaps, for example but I would like other people to elaborate this point.
- If the skin file contains blurry shapes, noises, etc ; you can safely reduce its size or use compression. I don't see the point of having 4k tyres skin files, for example.
- If the skin file have sharp shapes like inside banner, stickers, compression WILL produce artefacts. Are those artefacts noticeable or not depend on the source quality and the skin visiblilty from the player point of view. So, to me, carpaint dpesn't need to be uncompressed.
- If you decide to use compression, DXT1 produce the same quality than DXT5. The only difference rely on the loss of the alpha channel (if present). That is, you will have an on/off transparency with DXT1 while a graduated one with DXT5. To me, a carpaint without alpha channel can safely be saved as DXT1.
Finally, using compression or not has a simple answer : open the uncompressed file and compare it with the compressed one (at 100% zoom level)
A few (personal) tips from my days of skinning (gosh, it feels ancient said like this haha) :
- always remember the viewpoint : the size of the texture should always be equivalent to the size of the object in world space. As a reference, consider the "1 meter rule" : what is the relative size of the object from 1 meter afar, and what is your biggest texture size. For exemple, on a car with a full-shell skin, correctly mapped, at 4k, tyres/wheels shouldn't be above 1K, frontlights and rearlights above 512p, the largest interior map shouldn't be above 2k and everything "visible" only when closing on it between 512 and 128 maximum. The main reason is for coherence in pixel affectation : it looks better to have everything a tiny bit blurry than some things perfeclty sharp next to blurry things. keeping a "rule of size" is a great optimization trick, and helps a lot with visual logic. On the same idea, the viewpoint is also a question of "what is important". If you have to make compromise, you would prefer keeping a high res texture on the steering wheel, dashboard, and basically what's in front of your eyes while using the car than the seats, rear shell, and pretty much everything "out of view". It might look has important in showroom or in the making, but when optimization comes, only the viewpoint matter.
- No alpha = DXT1. Period. Like, really. Up to a third of the size for the exact same thing. Avoid mipmaps (useless for AC) and everything "un-necessary", like a big blank normal for a flat surface. If the normal is mandatory for the shader, a 16p is already too much
- Remember that the ratio of the texture matter, not the size. You can have a 1k texture/normal for the sharp tire for good rendering and a 512p texture and 256p normal for the blurred one, you probably wont see a difference in action.
Anyway, if you have any question, feel free to PM, i'll gladly help on what I can !