Hmmm.... Walt Disney just released their Ptex API ( http://ptex.us/overview.html )... perhaps I should add support for it in the 3.17.0...
It could be useful to get a super-detailed HP-baked base texture or to add micro-detail to the highpoly model.
Apologies if this has been covered before, but it's getting increasingly difficult to find answers to specific questions in a seventy-plus page thread....
I've started test baking for the portable comms unit I'm working on, and I've finished laying out the low-poly UV's as so:
So I tried a test bake in xNormal, with default settings and I end up with this normal map and the resulting artifacting:
I'm guessing that this is a result of how the mesh is constructed, or how I've set it up for baking? I've overlaid the low-poly model on top of the high-poly mesh and they occupy the same space in Max, and the error seems to be visible on the mesh using Xoliul's shader:
Is it best (or necessary) to set up the two meshes in Max as you would for an exploded bake before exporting the .obj files? Or is there something else I need to be doing to avoid this kind of artifacting? I'm still getting my head around the normal mapping workflow and I've only used xNormal for much simpler objects, so apologies if this has been covered or is painfully obvious.
Did you setup your cages or are you using constant ray distances?
That antenna can be considered as "floating geometry", so you should define a cage to limit the ray distances.... or explode the model.
Did you setup your cages or are you using constant ray distances?
I was using constant distances, but I suspect that the problem is some tiny pieces of geometry that weren't grouped with the other geometry, leaving the unaltered UV shells overlaying the rest of the UV layout, causing the artifacting.
Once that's fixed, I'll see how I get on with an exploded setup. Thanks for the tips!
hey all, was wondering if anyone could tell me why im getting some bad seems when baking in Xnormal? both shorts are from unreal, first image is with my lighting setup and all mats pulled in, the second is just the normal plugged into the diffuse spot so you can see the seems xnormal generates. My xnormal settings are the same as in the video metalliandy had linked to. thanks for any help you all can give
Can you post your UV layout and to select the conflictive faces there?
Probably there are UV seams in those faces. If not, maybe it's a problem with the smooth groups ( or the max2obj bugged exporter ).
hey thanks for a reply, yeah they are where the uv seems are. it happens where ever this is a uv border, is that what just happens or can i lay out uvs in a certain way to avoid it?
hey thanks for a reply, yeah they are where the uv seems are. it happens where ever this is a uv border, is that what just happens or can i lay out uvs in a certain way to avoid it?
If there is a UV seam the normals cannot be continuous, so the light neither will be.
Usually you want to hide those seams with hair or to put them in parts of the model which are hidden in some way ( for instance, at the bottom of the shoes, etc... ).
I'm afraid the UV layout needs some serious planning for normal and vector displacement maps :poly142:
Awesome update, Santy, this is going to speed things up immensely. Don't have anything on hand right now I could send you that's not under NDA though.
Bug: It seems the automatic naming for the resulting vertex color transfer map is using the _occlusion suffix, so it overwrites the ACTUAL ambient occlusion map that's been generated.
Bug: It seems the automatic naming for the resulting vertex color transfer map is using the _occlusion suffix, so it overwrites the ACTUAL ambient occlusion map that's been generated.
Yep.
Actually I found other bug: if you render an AO map and the highpoy mesh has vertex color the output will be the vertex colors, even if you check in the "Ignore vertex-colors" option.
3.17.0 Beta 2 released.
Try the Optix renderer! The speed is i.m.p.r.e.s.s.i.v.e
Just two notes:
1. The FW 196.21/196.34b crashes optix. You need the previous 195.62 ones.
2. Be sure you set the tile size to 256x256 ( or 128x128 if your card does not support it ) to use effectively all the parallelism the GPU can perform.
I'm using X-Normal, I've followed Matt Corcoran's tut and I'm still getting a lot of weird artifacts and other errata. I'm posting my views of the cage, the high poly and the low poly and what I'm getting out the other end of X-Normal.
Any ideas on why I'm getting problems?
Thanks for your help and advice.
Here is my UV and low-res:
The High-Res:
and the cage with the high-res superimposed:
And this is what I'm getting out. See all the strange colors? This is at 512, it's actually a lot clearer at 2048, but it's still got strange effects.
Hey Jogshy, I'm trying to use a .ctm file for my high res model in xnormal, but it doesn't show up as one of the supported formats. Did you take out the support for it?
Hey Jogshy, I'm trying to use a .ctm file for my high res model in xnormal, but it doesn't show up as one of the supported formats. Did you take out the support for it?
OpenCTM only works on x86 mode, not x64.
Also sure you're using the Beta 2, because the Beta 1 had a problem finding the the openctm.dll
@Tea Monster: Are you using cages? You should
If you used the lowpoly as basis for the HP's subdivision you could use the MatchUV feature.
Also an advice about your UV layout: keep some space near borders free for the dilation filter.... Btw, it seems the UVs on the bottom go behind the positive range?
Thanks, that did the trick. For some reason it didn't occur to me that it could have to do with x64. It's nice being able to use such small files. :poly121:
Im getting a number of cases where the occlusion renders out pure white or parts of a mesh are just rendered out white, using 3.17.0.41257. Normalmaps are fine. Ive not had this before 3.17, would this be due to the polypaint info or something? Im not exactly sure how to clear that information from a tool in zb, otherwise I would stick with 3.16
Im getting a number of cases where the occlusion renders out pure white or parts of a mesh are just rendered out white, using 3.17.0.41257. Normalmaps are fine. Ive not had this before 3.17, would this be due to the polypaint info or something? Im not exactly sure how to clear that information from a tool in zb, otherwise I would stick with 3.16
Play with the "Ignore per-vertex colors" option in the corresponding highpoly mesh's slot.
If a mesh has vertex colors exported and you don't check that option the AO will use the vertexColor.r component ( which is the same the Simple AO tool uses ). Checking the "Ignore per-vertex colors" forces xNormal to re-compute the AO using the software path.
I've got a problem with xNormal. I was using it yesterday (9th Feb) and it was working fine, I've gone to use it today and I keep getting an "expired" error and the program wont. (The filter in Photoshop also doesnt work) Ive tried reinstalling the program and this also doesnt work. Im using version 3.16.13 and I'm unable to download the new version as the links dont seem to be working. Can anyone send me the new version or help me fix the current version.
I've got a problem with xNormal. I was using it yesterday (9th Feb) and it was working fine, I've gone to use it today and I keep getting an "expired" error and the program wont. (The filter in Photoshop also doesnt work)
Yep, sorry. There was an error in the digital certificate and the computer's date. Workaround: alter the date of the computer.
The 3.16.13 is affected also but shouldn't produce the error until June 13.
I've run into a few potential bugs with xNormal 3.17 Beta 3 - 3.17.0.36069 x86.
The Optix renderer is FAST! Unfortunately it doesn't seem to obey the Max frontal or rear ray distance settings. It doesn't seem to cast any rear rays at all either.
Am I supposed to use a cage with the Optix renderer?
Here's a comparison:
I tried to use the Optix renderer to render a Normal and a Ambient Occlusion map at the same time by leaving them both checked "On" in the Baking options, however after clicking the Generate Maps button I got this error:
Then I unchecked Normal Map in the baking options, thinking that the Optix renderer might only be able to do one map at a time. Once I clicked the Generate Maps button I got this error:
Then my screen went black and came back on about a second later. Then Windows gave me this error:
At this point xNormal had to be shut down with the task manager, this also seemed to crash Modo which I had running at the same time.
Here's some of my system info, maybe it will help:
Windows Vista Ultimate 64bit SP2
Intel Core 2 Quad Q6600
8 GB system memory
GeForce 9600 GT 512 MB Ram
Driver version 195.62
Not related to those problems I get this error when I start up the x64 version of xNormal 3.17 Beta 3 - 3.17.0.36623 x64:
After I click "Ok" xNormal appears to startup and function just fine. I had the same error pop up with 3.17.0 Beta 2 but not with 3.16.12. The "cudart.dll" exists in both the x64 and x86 folders so I don't know what's happening there...
Anyway thanks for your support and development of xNormal
@Ben Apuna: I've just uploaded the beta3b solving some of the problems with the Optix renderer that you described. Pls, reboot the computer after installing the Beta 3b so the registry settings are updated correctly.
The "already mapped" problem that pops when you try to render two maps in one pass seems to be a NVIDIA bug though.
Exciting developments, Jogshy. Will the way you develop allow you to easily carry all of this work into Xn4?
I'm still very excited about the workflow enhancements that 4 will bring, and very glad you've decided to keep working on it!
Exciting developments, Jogshy. Will the way you develop allow you to easily carry all of this work into Xn4?
Well, initially I'm gonna use the CPU for all in xn4... and I'll incorporate all the GPGPU things once its basic structure is more or less stable ( with the expection of the GPU AO which will be present in the Alpha 1, yes or yes! )
xn4 is currently a bit halted because I'm waiting for a more stable LLVM version(for the programable shaders) and better drivers in linux.
I had also to re-design its SDK two times... but I think now it's fine!
Santy, did you stop taking donations? I can't see any PayPal donation button on the website any more, but you do still mention it in the FAQ:
Q: How much xNormal costs?
A: Absolutely nothing. It's free for any use, including commercial one. No price and no fees at all. If you like it, please, consider to make a donation to the help us or mention xNormal in your project.
cage vs no cage, im getting waves on my LP surface, so i went to no cage, and now im hitting a seam. I have all the uvw split by smooth groups and it is seam less with a cage just has wavy lines. but with out a cage it renders nicely with no waves but has seams why is this?
Jogshy, does installing Cuda 2.3 make any difference when using the Optix render?
Nope, I install all the required DLLs automatically... and, if, for example, you install CUDA 3.0 you shouldn't notice a significant improvement because the renderer is compiled offline using CUDA 2.3 so no optimizations could be applied just updating the CUDA runtime ( but perhaps the CPU-GPU transfers through the PCI express )
Remember you'll need the 195.62 drivers because the new ones don't work properly ( almost for Windows 7 ). For WinXP/Vista I have no idea :poly136:
Also remember you'll need the Beta 3b which solves a critical BSOD bug and disables the 5s watchdog properly for CUDA.
Oh... and you cannot render two consecutive maps... an "already mapped" error will pop ( I think it's an NVIDIA bug or I'm too blind to find my error ).
I was trying to implement the OpenCL version... but, unfortunately, the ATI OpenCL/OpenGL 3.2 implementation is still too immature to work... so I think I'll just postpone it until Q2 or 3.
Hey guys I was wondering if you guys knew what my problem was with this weird artifacing im getting on my Normal and AO maps. I can't figure it out and I need to get this piece done! If you could please help I would be in your debt!!
Hey guys I was wondering if you guys knew what my problem was with this weird artifacing im getting on my Normal and AO maps.
Need more info.
What artifacts are you referring to?
Are you using cages? Is the cage covering the highpoly mesh completely? Post all your settings, pls!
I was using the ray distance. I used the caculator to find the distance but it is still warping some of the normals and I can't figure out why. It also shows like a wrinkled effect on some parts.
Hi, a simple one this. My apologies, but I'm new to normal baking.
I've got a machine gun bullet box I'm doing. I've baked a high-poly bullet to a my LP one. I'm getting strange ribbed artifacts from the LP. What's the best way to get rid of these?
One thing I did do is to 'cheat' and I went into one of the channels of the map in Photoshop and put in a solid fill from the top to the bottom of the bullet. Is this 'kosher' or will it all end in tears?
cage - lp
These are the two normal maps. The 'ribbed' one and the one I fixed manually.
@Tea monster, Rebake with the low polygon's Smooth normals setting set to Average normals in xNormal (in the Low definition meshes tab) and see if that fixes it.
I get these issues if I've mistakenly exported my low poly mesh with hard edges.
I've got a machine gun bullet box I'm doing. I've baked a high-poly bullet to a my LP one. I'm getting strange ribbed artifacts from the LP. What's the best way to get rid of these?
What East said: try the "Average vertex normals" and see if that helps.
If not probably it's something related to the exported normals. I can see the face normals in that viewport but not the vertex ones!
I am getting a similar error message to the one Ben Apuna was getting. I have verified that I am using the correct Nvidia drivers and that I've restarted enough times for the registry settings in xNormal to be written. The error message I get is:
Memory allocation failed (Details: Function "rtContextLaunch2D"
caught exception: Encountered a CUDA error:
cuMemAlloc(&m_devicePtr, range_cast<unsigned int>(size)) returned
(2): Out of memory [7733306])
I've tried changing various settings in xNormal and none work. I'm 100% sure that I'm only trying to bake one map at a time because if I uncheck the one I am trying to do the "generate" button goes grey.
This is with the latest xNormal beta 3 that is on the site.
Memory allocation failed (Details: Function "rtContextLaunch2D"
caught exception: Encountered a CUDA error:
cuMemAlloc(&m_devicePtr, range_cast<unsigned int>(size)) returned
(2): Out of memory [7733306])
I've tried changing various settings in xNormal and none work. I'm 100% sure that I'm only trying to bake one map at a time because if I uncheck the one I am trying to do the "generate" button goes grey.
This is with the latest xNormal beta 3 that is on the site.
Sure you use the Beta 3b ( not the Beta 3 ) or you'll have some severe problems... ( /whistle whistle BSOD ) :poly136:
What ForceWare dirvers are you using? Seems Optix runs well in Windows 7 only with the FW 195.62 . The 196.21/192.34b are broken apparently.... I'm not sure about XP or Vista.
Well... that or your mesh does really need more VRAM
Btw, I think NVIDIA gonna release a new version of Optix this Friday so perhaps I could post a new release soon
Replies
http://downforeveryoneorjustme.com/eat3d.com
Try clearing your cache and see if that helps
It could be useful to get a super-detailed HP-baked base texture or to add micro-detail to the highpoly model.
I've started test baking for the portable comms unit I'm working on, and I've finished laying out the low-poly UV's as so:
So I tried a test bake in xNormal, with default settings and I end up with this normal map and the resulting artifacting:
I'm guessing that this is a result of how the mesh is constructed, or how I've set it up for baking? I've overlaid the low-poly model on top of the high-poly mesh and they occupy the same space in Max, and the error seems to be visible on the mesh using Xoliul's shader:
Is it best (or necessary) to set up the two meshes in Max as you would for an exploded bake before exporting the .obj files? Or is there something else I need to be doing to avoid this kind of artifacting? I'm still getting my head around the normal mapping workflow and I've only used xNormal for much simpler objects, so apologies if this has been covered or is painfully obvious.
Thanks in advance for any suggestions or input.
That antenna can be considered as "floating geometry", so you should define a cage to limit the ray distances.... or explode the model.
I was using constant distances, but I suspect that the problem is some tiny pieces of geometry that weren't grouped with the other geometry, leaving the unaltered UV shells overlaying the rest of the UV layout, causing the artifacting.
Once that's fixed, I'll see how I get on with an exploded setup. Thanks for the tips!
Probably there are UV seams in those faces. If not, maybe it's a problem with the smooth groups ( or the max2obj bugged exporter ).
Usually you want to hide those seams with hair or to put them in parts of the model which are hidden in some way ( for instance, at the bottom of the shoes, etc... ).
I'm afraid the UV layout needs some serious planning for normal and vector displacement maps :poly142:
For the Beta 1 I'll add some Optix-accelerated ray tracing things
Btw... anybody wanna send me a ZB3.5 Polypainted(MRGB/RGB vertex colors) model to add to the xNormal's examples, pls?
Bug: It seems the automatic naming for the resulting vertex color transfer map is using the _occlusion suffix, so it overwrites the ACTUAL ambient occlusion map that's been generated.
Actually I found other bug: if you render an AO map and the highpoy mesh has vertex color the output will be the vertex colors, even if you check in the "Ignore vertex-colors" option.
Bugs are funny!
Try the Optix renderer! The speed is i.m.p.r.e.s.s.i.v.e
Just two notes:
1. The FW 196.21/196.34b crashes optix. You need the previous 195.62 ones.
2. Be sure you set the tile size to 256x256 ( or 128x128 if your card does not support it ) to use effectively all the parallelism the GPU can perform.
I'm using X-Normal, I've followed Matt Corcoran's tut and I'm still getting a lot of weird artifacts and other errata. I'm posting my views of the cage, the high poly and the low poly and what I'm getting out the other end of X-Normal.
Any ideas on why I'm getting problems?
Thanks for your help and advice.
Here is my UV and low-res:
The High-Res:
and the cage with the high-res superimposed:
And this is what I'm getting out. See all the strange colors? This is at 512, it's actually a lot clearer at 2048, but it's still got strange effects.
Also sure you're using the Beta 2, because the Beta 1 had a problem finding the the openctm.dll
@Tea Monster: Are you using cages? You should
If you used the lowpoly as basis for the HP's subdivision you could use the MatchUV feature.
Also an advice about your UV layout: keep some space near borders free for the dilation filter.... Btw, it seems the UVs on the bottom go behind the positive range?
I got the errors on the bottom when I calculated the distance of the raycasting. When I set it to default, it worked fine.
Thanks again for this wonderful software!
Optix, OpenCL, OpenGL 3.2 and DX11 are distracting me too much ! :poly136:
If a mesh has vertex colors exported and you don't check that option the AO will use the vertexColor.r component ( which is the same the Simple AO tool uses ). Checking the "Ignore per-vertex colors" forces xNormal to re-compute the AO using the software path.
I've got a problem with xNormal. I was using it yesterday (9th Feb) and it was working fine, I've gone to use it today and I keep getting an "expired" error and the program wont. (The filter in Photoshop also doesnt work) Ive tried reinstalling the program and this also doesnt work. Im using version 3.16.13 and I'm unable to download the new version as the links dont seem to be working. Can anyone send me the new version or help me fix the current version.
Thanks in advance
The 3.16.13 is affected also but shouldn't produce the error until June 13.
The current 3.17b3 has that bug fixed.
I've run into a few potential bugs with xNormal 3.17 Beta 3 - 3.17.0.36069 x86.
The Optix renderer is FAST! Unfortunately it doesn't seem to obey the Max frontal or rear ray distance settings. It doesn't seem to cast any rear rays at all either.
Am I supposed to use a cage with the Optix renderer?
Here's a comparison:
I tried to use the Optix renderer to render a Normal and a Ambient Occlusion map at the same time by leaving them both checked "On" in the Baking options, however after clicking the Generate Maps button I got this error:
Then I unchecked Normal Map in the baking options, thinking that the Optix renderer might only be able to do one map at a time. Once I clicked the Generate Maps button I got this error:
Then my screen went black and came back on about a second later. Then Windows gave me this error:
At this point xNormal had to be shut down with the task manager, this also seemed to crash Modo which I had running at the same time.
Here's some of my system info, maybe it will help:
Windows Vista Ultimate 64bit SP2
Intel Core 2 Quad Q6600
8 GB system memory
GeForce 9600 GT 512 MB Ram
Driver version 195.62
Not related to those problems I get this error when I start up the x64 version of xNormal 3.17 Beta 3 - 3.17.0.36623 x64:
After I click "Ok" xNormal appears to startup and function just fine. I had the same error pop up with 3.17.0 Beta 2 but not with 3.16.12. The "cudart.dll" exists in both the x64 and x86 folders so I don't know what's happening there...
Anyway thanks for your support and development of xNormal
The "already mapped" problem that pops when you try to render two maps in one pass seems to be a NVIDIA bug though.
I'm still very excited about the workflow enhancements that 4 will bring, and very glad you've decided to keep working on it!
xn4 is currently a bit halted because I'm waiting for a more stable LLVM version(for the programable shaders) and better drivers in linux.
I had also to re-design its SDK two times... but I think now it's fine!
CAGE:
NO CAGE:
http://www.poopinmymouth.com/tutorial/normal_workflow_2.htm
See the cylinder.
Remember you'll need the 195.62 drivers because the new ones don't work properly ( almost for Windows 7 ). For WinXP/Vista I have no idea :poly136:
Also remember you'll need the Beta 3b which solves a critical BSOD bug and disables the 5s watchdog properly for CUDA.
Oh... and you cannot render two consecutive maps... an "already mapped" error will pop ( I think it's an NVIDIA bug or I'm too blind to find my error ).
I was trying to implement the OpenCL version... but, unfortunately, the ATI OpenCL/OpenGL 3.2 implementation is still too immature to work... so I think I'll just postpone it until Q2 or 3.
Thank you!
What artifacts are you referring to?
Are you using cages? Is the cage covering the highpoly mesh completely? Post all your settings, pls!
I've got a machine gun bullet box I'm doing. I've baked a high-poly bullet to a my LP one. I'm getting strange ribbed artifacts from the LP. What's the best way to get rid of these?
One thing I did do is to 'cheat' and I went into one of the channels of the map in Photoshop and put in a solid fill from the top to the bottom of the bullet. Is this 'kosher' or will it all end in tears?
cage - lp
These are the two normal maps. The 'ribbed' one and the one I fixed manually.
I get these issues if I've mistakenly exported my low poly mesh with hard edges.
What East said: try the "Average vertex normals" and see if that helps.
If not probably it's something related to the exported normals. I can see the face normals in that viewport but not the vertex ones!
I'm still not quite getting the hang of the ray calculator though. I followed the directions and I still get 'drop outs'. Any ideas on that?
Thanks again for all your help.
I've tried changing various settings in xNormal and none work. I'm 100% sure that I'm only trying to bake one map at a time because if I uncheck the one I am trying to do the "generate" button goes grey.
This is with the latest xNormal beta 3 that is on the site.
Sure you use the Beta 3b ( not the Beta 3 ) or you'll have some severe problems... ( /whistle whistle BSOD ) :poly136:
What ForceWare dirvers are you using? Seems Optix runs well in Windows 7 only with the FW 195.62 . The 196.21/192.34b are broken apparently.... I'm not sure about XP or Vista.
Well... that or your mesh does really need more VRAM
Btw, I think NVIDIA gonna release a new version of Optix this Friday so perhaps I could post a new release soon