Another problem baking on xnormal ...
polycounter lvl 14
Hello , I have made what I think should be a pretty simple shape ... you can see here a topogun screenshot of the high model reference and above the lowpoly that fits perfectly on the high poly , the general wall baking comes out well but the windows intrusions do not so I detached in uvmapping and pelted and relaxed placing them at the corners of the textures , they are even and uniform in scale with the wall etc ... but the problem is that in xnormal , no matter whatsettings I put for the ray distances , it always make the sloppy results you can see in the normal map , just on the windows , now why that? the low poly follows precisely the shape inside the intrusions , the uvs are even and placed well ... why it does mess it up ? I have tried ray distances for front and rear from 2 up to 0.001 and no improvements , I mean sure it changes but not significantly and still it offers bad results....
Then I tried with a cage and the result are near to awfull as the whole wall gets ok only if I enlarge the cage but then the windows get fagocitated in vertex and faces intrusions ...
What can I do to bake this simple shape that is driving me mad now?
Then I tried with a cage and the result are near to awfull as the whole wall gets ok only if I enlarge the cage but then the windows get fagocitated in vertex and faces intrusions ...
What can I do to bake this simple shape that is driving me mad now?
Replies
Edit:Unless you are moving verts, uv's or the cage.
Edit2: Oh and this method won't work with a custom cage from Xnormal as the topology changes.
I am starting to think it might be easier to just diagnose your cage issue in Xnormal
I would have to see a screen of your cage in Xnormals model viewer, preferably at an angle where I can see the window well, question, are you clicking save models before exiting Xnormal's 3d viewer?
http://www.speedyshare.com/FUNUX/Nordic-Mound.7z
Now, sorry to come off like a dick, but the low poly is fucking appauling. You have access to topogun, so why not use their great retopo tools? It seems like you just decimated a mesh in zbrush and call that your low poly and called it a day.
I do not know much topogun as I am tring it for first times and I used only create the vertex , and faces and i have united them all without caring much for quads as far as I know for a normal static mesh is not needed . so what's the not good?
Also the texture in xnromal is enlarged to a 2014x1024 but if you open in max it looks stretched just because I can't make a uvmap non squared in 3dsmax so did stretch ...
What mens appauling? and how shoudl be done to be not appauling?
so can u tell me how to fix?
Also try turning off the ray miss option.
then I 'd still like to knw why the uvmapping is not good .
If there are overlapping portions then bakes won't turn outright, also the edges have lots of space in your UVs but the details near the center have barely any space, also it's extremely stretched. Normally before you do bakes and when you unwrap, you should throw a checker pattern of some sort onto your mesh to see where there is stretching. You normally want to have as little stretching as possible on your meshes.
Even if you mesh is static and all your textures are only going to be baked I feel like having less stretched textures will make things easier if you ever need to paint into your diffuse texture and it helps the texture density be more even throughout the entire model.
Something else that may cause problems with the results of bakes if your topology I think, you have lots of very narrow triangles, I believe that can cause problems when baking some of your maps.
Here is how the Uvmapping and distribution of checker pattern looks in3dsmax and with the non even quadlike but rectangular uvmapping ... is it bad? If so I do not get why because all is even and all has space for the texture , so what's the problem of that uvmapping?
the uvmap above was produced with a quad layout but mine is rectangular ...
here they are ...
what's wrong with those? why I get the artifacts if the low poly just borders, the high poly ? there are no intrusions or weird parts on the intrusion windows surfaces etc...
some possible reasons
Ray is going too far
Ray is not going far enough
Ray is being sent off an an unintuitive angle
Ray is hitting some unseen or backfacing geometry and terminating
If you are changing the ray distance between high and low values you might be swapping between the first two and so need to use a cage to have variable ray lengths.
I don't think your UV or low poly model are causing the errors unless you have something like overlapping intersecting or non manifold geo. (I haven't looked)
Look at the image you posted and say what you think the problem is.
I couln't find any valid reason why those bad artefacts did appear , I checked all normals and to me they looked correctly extruding from the vertices and so rays should hid just the mesh and not other things ... is also working fine on the even surfaces but not just on the intruded windows ....
I cleaned up the UVs a bit, moved them around (not perfectly) but the UVs are more evenly distributed now.
bake in xnormal, viewed with standard phong realtime view in Maya 2011,
and the normalmap,
I expect as you changed the ray distance from 0.001 to 2 you went from having areas of short rays terminating too early and producing blank areas to having long rays not terminating early enough and hitting geometry you don't want.
It sounds like you're not thinking the process though clearly enough, try and visualize the rays emitting from each pixel and where do they go and what makes them stop to produce the final result.
you should extract a problem areas and test with only them, I'm expecting a cage to be the only solution. The cage is a mess but it is still representative of what the bake is doing, then you only need to go in and fix the problem areas by making sure the cage isn't overlapping unwanted geo.
That would explain why your 2 days of ray length tweaking didn't significantly improve the bake. Because there are 2 different errors breaking the bake and the model is one you cant bake with a single ray length so you will never get a clean bake in that way.
I don't think you know how to use a cage properly, explain how you think it works and what problems it solves.
I did notice you have some uv faces that are welded wrong. In one of the previous posts of the UV editor, you can see them as marked red. These arent overlapping, but they arent orientated correctly and one of them is welded to the wrong vertices and should be corrected. (Mirror on X or Y, or do a Smooth UV on that island.)
I used the raydistance calculator in xnormal to get the ray distance,
Normalmap settings are
"Closest hit if ray fails" unchecked.
"Discard Back-faces hit" checked.
let me know if you want me to upload the OBJ of the changes i made.
1.) all fully automated bake > some errors in cage projection along wall openings. (this would for sure require little love by hand adjusting cage. can be done either in your 3d app or in xnormal)
2.) your UVs seem to have some extra splits around outer edge.
3.) triangulate before toying with cage and bake (i think you already know that)
4.) if all fails use ray distance calculator which should be located in tools of xNormal
Even you just inflate the low poly model by x amount it's still an representation of how the normal map baker does it behind the scenes with the ray distance also set to x.
There's no way you should need to spend 2 days testing ray amounts from 0.001 to 30. The ray distance is a unit of measurement that can be directly related to your model (although I say that, it could be a % of the models bounding box)
I guess I need to elaborate, you see the red parts? they are flipped or backfacing UV's, it's fine to have them as long as the entire Island/Cluster is one or the other, you can't have both on the same island, I believe this is half of your problem.
Oh and you have some uv overlap in the bottom right corner, this obviously won't work either.
I used the ray distance calculator and I get those results ...
LP_.obj fmin 0 fmax 97 rmin 0 rmax 354
here is the result , I guess a little better but still there are some issues as you can see ... btw the uvmap of the model I am using is a little bit different from the one u have downloaded because I fixed the issues that where rised up b4 ....
as you can see I still having those weird strange linees etc ... what can those be due to and how to remove definetly ?
http://dl.dropbox.com/u/2516103/LP_Nordicmound_UV.rar
https://dl.dropbox.com/u/1693140/web/Untitled-5.jpg
https://dl.dropbox.com/u/1693140/web/Untitled-3.jpg
Steps you should follow to making a cage, it wont take a week, it shouldn't even take more than 10 mins for something like this.
Make "ugly" looking cage as you have already done by pushing it out an even and reasonable amount.
Test bake, it should look the same as a normal bake without the cage
Find all the errors, go in and move the cage so its not intersecting unwanted geometry.
Test again.
Done or fix more errors.
Actually I'm pretty lazy so I would just flatten the tight indents where the problems are and it will be ok.
but the main problems I am having are just on the wndowed parts ...
what u mea by flatten tight indents?
The problems I am referring mostly are in the larger intrusion windows walls .
This is an example of the first problem, its a slice of the the white lines are uniform ray distances and the red circles are the geometry the bake is deciding to use the normal from.
https://dl.dropbox.com/u/1693140/web/Untitled-2.jpg
Yeah this is really good advice, just keep your low simple and try not to cut in little gaps like this that are going to cause errors when the cage/rays get pushed out and overlap.
while loading a high res poly in 3dsmax , well it slows a lot my system ....
Don't rush things - just take the time to do things well.