Normal map seams on UV borders

vempdance

Sorry for bothering you with this simple question.

When I bake normal map use xNormal,I got seams on every uv borders.

I don't use any mirrored uv shells, and I have enough edge paddings. The direction of RGB channels have set correctly. And the map is linear color space.

What else could cause this problem and How to fix it?

ZombieWells

here is everything you need to know. Watch the video, you have your issue and solve there. Then read everything else! http://www.polycount.com/forum/showthread.php?t=107196

vempdance

ZombieWells wrote: »

here is everything you need to know. Watch the video, you have your issue and solve there. Then read everything else! http://www.polycount.com/forum/showthread.php?t=107196

Thanks for the reply.
In fact I have already read this thread and it dosen't solve my problem. And I also googled a lot to find the solution, but all of them are the situation that I write in the first post and does not match mine.So I ask here.
Thank you again.:)

peanut™

From what im reading have you tried to bakejump from an other platform ?

not that it's unthinkable but if padding is enable and creates problem try from 3dsmax or maya instead of xnormal and see if there's some improvement. At least that's what i would do.

AlecMoody

You should post pictures. Having an extremely subtle seam where you have a split is really common but in practice it shouldn't be visible. If you have something more than that there is a workflow issue.

vempdance

peanut™ wrote: »

From what im reading have you tried to bakejump from an other platform ?

not that it's unthinkable but if padding is enable and creates problem try from 3dsmax or maya instead of xnormal and see if there's some improvement. At least that's what i would do.

Ah，my computer is not so good so loading a high polymesh into maya always make her refuse to work:(...
Thanks:thumbup:

vempdance

AlecMoody wrote: »

You should post pictures. Having an extremely subtle seam where you have a split is really common but in practice it shouldn't be visible. If you have something more than that there is a workflow issue.

Sure, I should have post the picture.Sorry for not doing that.

It is very little, and won't be recognized until I zoomed in a lot.
The next picture is from a higher distance.Sorry that I don't draw the red circle at the right position, the seam is at the very right inside the red circle.


And this is what normal map looks like in the border place.(At the top of the UVed place)


Is this what you mean "subtle seam"? Is it means that this can't be solved in theory, but it is very difficult to be recognized in prectice so we just ignore it?

Zezeri

Do you have hard edges on all of your UV Borders?
Also, show pictures

Where do you view the normals?

EarthQuake

Generally, there will always be a seam if you zoom in far enough. This looks like you're zoomed in like 50x farther than you should be. What does the full asset look like and what is the intended viewing distance?

vempdance

Zezeri wrote: »

Do you have hard edges on all of your UV Borders?
Also, show pictures

Where do you view the normals?

Yes I have tried that and doesn't work.

I view it in marmoset toolbag 2.

Thanks

vempdance

EarthQuake wrote: »

Generally, there will always be a seam if you zoom in far enough. This looks like you're zoomed in like 50x farther than you should be. What does the full asset look like and what is the intended viewing distance?

This is the picture from a higher distance. It seems a new lighting error appears.

The seam is at the right side of the light area, the left side does not have a seam.

EarthQuake

In TB2, make sure to set the tangent space in mesh properties to where you baked the map, in this case xnormal/mikktspace

vempdance

EarthQuake wrote: »

In TB2, make sure to set the tangent space in mesh properties to where you baked the map, in this case xnormal/mikktspace

Yes, I've already set that.

EarthQuake

Not sure if this is your final uv layout, and it probably won't help your seam issue, but if you clamp the edges of your uvs to be straight rather than curved the normal map will look a lot better, you'll get rid of all those aliased edges. You only need 1 row of pixels to draw a straight line, but an angled or curved line needs at least two.

vempdance

EarthQuake wrote: »

Not sure if this is your final uv layout, and it probably won't help your seam issue, but if you clamp the edges of your uvs to be straight rather than curved the normal map will look a lot better, you'll get rid of all those aliased edges. You only need 1 row of pixels to draw a straight line, but an angled or curved line needs at least two.

I tried that, and yes it doesn't help:poly124:.
Thanks.

WarrenM

Well, SOMETHING must be wrong, dude. I mean ... if you've tried everything suggested above and it all checks out then I think you've reached the limits of what the internet can offer.

AlecMoody

Best debug practice would be to bake to object space and evaluate your object space normals on the model. If the problem persists, you know you have a projection issue. If the problem goes away, the issue is related to tangent space, precision, or how your mesh is set up.

vempdance

AlecMoody wrote: »

Best debug practice would be to bake to object space and evaluate your object space normals on the model. If the problem persists, you know you have a projection issue. If the problem goes away, the issue is related to tangent space, precision, or how your mesh is set up.

The object sapce normal map still shows the seam.

my work flow of baking is like this(in maya):
1.I triangulared my low mesh,made all edges to soft edge,and exported it out.
2.Load high mesh as a reference, edited the low mesh vetexs to cover the high mesh, then exported it as cage.
3.Load high mesh,low mesh and cage into xNormal, set mesh scale to 10, map size to 2K, edge padding to 16, then use OpenRL renderer to bake.(I aslo tried the default bucket renderer but get the same resault.)

What may have caused the projection issue?

Neox

the first thing to lift the quality up would be to straighten those UVs that small seam is a smaller issue than all those pixels you can clearly see

vempdance

Neox wrote: »

the first thing to lift the quality up would be to straighten those UVs that small seam is a smaller issue than all those pixels you can clearly see

Ah, that is because these are test bakings so I don't care of it.When come to the final maps I will turn on the antialising to get a better looking.
Thanks.

AlecMoody

If you want, you can email me an FBX of your lowpoly model and your object space normals and I can tell you why this is happening. The highpoly for that specific section might also help.
alecmoody@gmail.com

vempdance

AlecMoody wrote: »

If you want, you can email me an FBX of your lowpoly model and your object space normals and I can tell you why this is happening. The highpoly for that specific section might also help.
alecmoody@gmail.com

I've already send a email.
Thanks!:thumbup: