Ah, guess I should have clarified in my mini tut: the point of splitting the UVs is to distance them so that the border edges don't land on the same pixel of texture. Give them the same distance you would between any standard UV island.
A V G The following art and other content is from a game project I've been occupied with for about 4-5 years now. In short, the game is antinatalism themed and has a quite pessimistic story. It would have various settings, from Earth in the near future (not shown in this post) to various alien planets and megastructures in…
yeah. I agree it looks like a cool idea. Still I am not sure , for final result it goes as modified binary mask still for alpha test only or they render actual translucent pixels on the edges? Just a few on the very edge? So another render pass / buffer image?
"Build out the bushes, the foreground, the sky.." HAHAHAHAHAHAH! That a genius boy! This seems so.. illogical to me? Its one thing to identify the surrounding pixels and make a well calculated guess - which I suppose, couldn't be achieved so accurately with computing - and another to guess the ELEMENTS surrounding other…
The edge "fuzziness" is called edge padding/dilation: http://wiki.polycount.com/wiki/Edge_padding It helps reducing bleeding of pixels to neighbour uv shells when the texture gets mipped (lod-d). What do you mean by "format"? Do you refer to the way things are arranged and combined together?
Thx, entering the precise value while constraining is a good solution. Just to re-iterate - with UVs -> Snaps to Pixel enabled - move your shell to a corner - g to translate - press x/y to constraint - enter number to translate with precise value. (also you can use - for negative number)
So confused a = no normal 'per vertex' b = normal, 'per pixel' Surely b is correct? When I see correct curvatures they have lighter tones towards edges,makes sense. So whilst method a fixes the glitches I don't get the white lines. Can't win.
You can't delete topology after the bake when using tangent space normal maps, as the data is based on the surface/vertex normals. Even if you spin an edge it will disrupt the pixel information. This can be done with an object space map and converted to a TS map after the fact. It's called skewmeshing.
thx pyrzern ... looking cool man .. well distributed pixels there jimmyrustler : thx man . im flattered :poly136: of course feel free . it still needs adjustments ... right now the belt can easily fall off the iron piece haha need to fix that . glad ur joining BTW
Those are smoothing groups. The shading is per-vertex. Add more geometry, and they won't be as bad. When you render, the shading is per-pixel I think. Turbosmooth the model, you'll see, you won't see any more "strange dark lines", which aren't strange at all. It's normal.