bad bake seams?

jordank95

Trying to get some bakes done. This is just a simple 6 sided mesh piece. All UV shells are their own shell since it's 90 degrees and each UV shell is hardened. Im still getting these noticeable seams at all corners. It looks fine from a little zoomed out, but is this normal? How do I fix this?

Using Maya and Substance Painter.

Zoddo

I had this same issue too, look at these threads:

http://polycount.com/discussion/comment/2631386#Comment_2631386
http://polycount.com/discussion/200700/substance-painter-baking-problems-seams-seams-everywhere#latest

Long story short, problem might be with your smoothing groups or edge loops.

jordank95

Zoddo said:

I had this same issue too, look at these threads:

http://polycount.com/discussion/comment/2631386#Comment_2631386
http://polycount.com/discussion/200700/substance-painter-baking-problems-seams-seams-everywhere#latest

Long story short, problem might be with your smoothing groups or edge loops.

Thanks. Seems spacing my UV shells a little more apart seemed to help, though its not fully fixed yet. Not sure what else to do. All those pieces are their own UV island.

Can anyone explain what the Dilation Width is and what it does? Also, is there a way to tell how many pixels apart my UV shells are? Is there a script to space them all apart at a specified pixel distance?

Mink

Post your UV layout, preferably with some sort of color grid for reference. It looks like you're trying to make the edges of your chair seamless, which usually results in normal seams when baking. I'm not sure what you're using, but for even UV island spacing I use TexTools for blender.

jordank95

Mink said:

Post your UV layout, preferably with some sort of color grid for reference. It looks like you're trying to make the edges of your chair seamless, which usually results in normal seams when baking. I'm not sure what you're using, but for even UV island spacing I use TexTools for blender.

Not at my computer currently - but any face thats 90 degrees or greater, I split into its own UV island. So, no it's not seamless. Each one of those sides is its own UV island. So 6 total UV islands for that piece in the photo thats causing the issues.

Mink

Are you smoothing your lowpoly by UV shells?

jordank95

Mink said:

Are you smoothing your lowpoly by UV shells?

Every UV shell is hardened. I have a script that allows me to select all my UV shells and harden edges according to the UV shells.

Kanni3d

Hard to tell unless you show us your UV's when you're available. Looks like there's more to the mesh from the screenshot, and this simple can be a low texel-density issue

jordank95

Kanni3d said:

Hard to tell unless you show us your UV's when you're available. Looks like there's more to the mesh from the screenshot, and this simple can be a low texel-density issue

Heres the UVs. All pieces have even texel density. Baking 2k normal maps. I've darkened the UV shells that are the ones in the picture above. 



I adjusted the UVs a little bit and set the Dilation wide to 16 and got a little better result, but its not perfect. Here's the results:



Could it be that I'm just looking at mesh too closely and nitpicking? You will never see the mesh this close up and once all maps are on. By the way, this is an old TV, so that piece is a side wood piece.

Kanni3d

jordank95 said:

Kanni3d said:

Hard to tell unless you show us your UV's when you're available. Looks like there's more to the mesh from the screenshot, and this simple can be a low texel-density issue

All pieces have even texel density. Baking 2k



Could it be that I'm just looking at mesh too closely and nitpicking? You will never see the mesh this close up and once all maps are on. By the way, this is an old TV, so that piece is a side wood piece.

Even or not, I just meant the shells themselves were probably not receiving a whole bunch. There's a couple things here, the UV's have a lot of wasted space, especially since they're all straight/rectangular uv's, you should get a very tight pack out of it.  What's your high-poly look like?

But yes, there comes some times where you're simply looking too close and being nitpicky, it happens

guitarguy00

I also encounter this problem sometimes and can't figure it out. It's only when up close and is more evident when I turn the gloss up in Marmoset but all my UV shells have their own smoothing group and I make sure to have a UV seam where there are angles above 70 degrees. I even straighten all the UVs that have curves in them(when appropriate). Quite strange.

poopipe

jordank95 said:

Can anyone explain what the Dilation Width is and what it does? Also, is there a way to tell how many pixels apart my UV shells are? Is there a script to space them all apart at a specified pixel distance?

Dilation is where the pixels at the edge of your uv shell are stretched outside the shell,  dilation distance is how many pixels  that goes on for.

Dilation is there to help deal with filtering and mipping artefacts  and is absolutely essential if your UVs aren't snapped to pixel boundaries. 

strictly speaking at 2k you need a minimum of 8 pixels dilation to prevent mipping artefacts which means you want a total of 16 pixels between shells.  In practice you can get away with half that for most hardsurface type assets. 

The distance you need is connected to texture size due to the way mip-mapping works - i believe there's an entry in the polycount wiki explaining it all 


There is an element of nitpicking about this - normal maps are not perfect and this is the worst case scenario for a normal map bake,

If you need to see the edge of your asset that close and it's supposed to be bevelled you're much better off actually bevelling it. 

Mehran Khan

I ran into similar problem, the answer is that there is no solution to this. I tried everything! from uv corrections, to making a cage . Having a cage reduces this alot but it does not go away at zoom levels.
Someone recommended me to bake inside of substance designer as this is seemingly a bug inside of painter, but I did not try that.

SebastianBielecki

Ha... this is interesting, how it looks on normal map?

Zodd

Yeah, everything is ok with the hard edges and split uv islands, there should be no need to be specific on how many pixels is the spacing between the islands.
here is something that may work (up to varying degree of success because... normal maps
Try baking normal map in maya if it bakes out fine (ie no seams) its the problem with normal map/tangent space calculating algorithm thingy, in short different bakers have different results (slightly but enough to cause seams ) so if you have good results from maya you
either bake in substance entirely, or try triangulating low poly>freeze normals>bake in maya and use that normal map in substance.

cheers and happy baking
edit:
regarding normal map spaces and thingys, this may be usefull:
https://forum.allegorithmic.com/index.php?topic=10304.0