seamless tiling of UV's

Graphics and GPU Programming Programming

Started by proanim October 23, 2013 04:45 PM

1 comment, last by proanim 10 years, 5 months ago

proanim

455

Author

October 23, 2013 04:45 PM

Can someone explain what is 'the most optimal' way of tiling uv coordinates on a model?

I wan't this to be reliable enough to not produce seams or artifacts of any kind since all my models will use diffuse/normal/specular maps. I do know UDK and id tech has this in more complicated/advanced way, and I think that it might be overkill for me to do it that way.

So I was thinking of simply tiling uv's every time when object is scaled, for example 2x, 3x, 4x, etc. UV's would be tiled for number of times object was scaled compared to it's original size.

Krohm

5,051

October 24, 2013 07:51 AM

I always built uvs by just using a thing like that:

(u,v) = f(x, y, z)

Yes, pure function. It will work under the following assumptions:

Your vertex values are unique (shared vertices collapsed);
No "spurious" vertices;
Texture actually tiling.

So, if you have "seams or artifacts" I'm inclined to believe at least one of the preceding requirements are not met.

It would be nice to know what those "seams or artifacts" look like as a start!

I think the most common problem is missing (1). This will cause small floating point inaccuracies at best and just different results in other cases. Are you 100% sure the input dataset is correct?

Previously "Krohm"

proanim

455

Author

October 24, 2013 07:22 PM

Ok thanks, it took me awhile to find out that my math was slightly off, since i also used what is practically the same function.