[ QUOTE ]
In the ambient occlusion options, what is the Bias value for? I don't see an explanation for it in the help file.
[/ QUOTE ]
It basically extrudes a bit the highpoly hit point to solve some self-occlusion problems. It works in the same way that Blender... see this image:
Bias 0.0001:
Notice the self occlusion problem at the triangle edges...
If you increment the AO bias a few to 0.015:
Also, playing with the AO angle aperture degrees you can get a dirtier AO map or cleaner(or touch the AO method for a more-drastic result.. from cosine to uniform, for example)
And remember... you can(optionally) use the cage/maximum uniform front/back ray distance to control the ray extension longitude... or uncheck the "don't limit ray distance" to disable it.
Also remember you don't need a highpoly+lowpoly model to render the AO. You can render the AO with just one model(selecting it as lowpoly and highpoly, the same file in the slots). If you use this, sure you set well the cage+a greater AO bias or you will get a void AO map!(because rays won't collide themselves).
You could use too the new render fast AO with model+normal map that Jonhy luvs!(hope)
ps: Omg Fat, I saw that nice hair video in your web and now I can't stop thinking about how to implement it in realtime! Why why why you put there!!! Now I can't sleep!
You mean the hair on the guy? I did that with HairFX in Max. If you can get that in real-time, you'll have a lot of people knocking on your door.
And I noticed that a couple people here have said that the ambient occlusion from xNormal isn't as good as other packages, so I did some tests. For some reason you don't seem to get the detail in xNormal that you do in Max, for example. Here's a picture...
I tried to get the settings between the Mental Ray AO shader and xNormal's AO as close as possible. But you can see that there's more detail in the MR rip. I couldn't say why, but I think that's what people are talking about. Any thoughts on where the difference is coming from?
[ QUOTE ]
You mean the hair on the guy? I did that with HairFX in Max. If you can get that in real-time, you'll have a lot of people knocking on your door.
[/ QUOTE ]
Yep, was that Amazing!
[ QUOTE ]
For some reason you don't seem to get the detail in xNormal that you do in Max, for example. Here's a picture...
[/ QUOTE ]
Looks like you need to play a bit with the Ao angle(tru to increase or decrease it). Play too a bit with min/max output levels, the right image looks a bit saturated.
Btw, caution if you are saving to OpenEXR file format ( need to set the exposure and gamma well when importing/vieweing the file in Photoshop, etc ) or the white color will "saturate"
Also try to reduce the AO bias(but I think the default is the minimum...). Try always to put it to zero, unless artifacts appear(like in the Blender ball I posted before).
Also very important, set the minimum #of rays and maxiumum EQUAL ( so adaptative AO will be disabled ).
Also sure don't use the black sand median filter neither the blur filter.
Finally, try to scale the mesh up. To avoid small floating point errors I extrude a constant distance(something like 0.000001f) the ao rays origin so scaling the mesh a bit up should help.
But well... to get the desired microdetail there probably you should use a cavity map(or a thing I have here prepared for the next version )
And ... MR uses area-51 technology(deepshadow volexized rays,energy radiance dispersion with BDRF,128bits floating point precision, micro-patches ala REYES,blah blah) I can't barely dream So if I can aproximate, just a bit, to its results I would be happy!
Well, let me revise a few the code if I can find something wrong. Actually I think I know what's happening( I made a mistake in the AO formulae, but i'm not completely sure... let me make some tests).
Great discussion on the AO. I too was a little disappointed about getting good results with the AO. I have been multiplying my cavity map over my AO to get a more detailed look.
Feature Idea - It would be nice, and sorry if it exists and I can't figure out a way to do it, but to have hi-res meshes exist only to cast "shadows" in the AO computation. These meshes are not calculated in any other way, they don't get hit with normal rays, they don't get hit with Ao rays, etc.
Why would one need this? Example. I have a character with a great deal of armor collar around his neck. This collar needs to be a separate model. I now want to calcualate the head and neck AO but the collar is so close there is no way for me to not get eithe rmy cage or rays to not intersect it in some way. Also if I use rays it is much easier to process the head on it's own pass. Now if I could load the collar up to "Cast AO ONLY" it would be almost as if it did not exist in the calculations except to cast shadows on the hi poly head that are then calcualted in the heads AO map. I hope that makes some sense?
Feature Idea - It would be nice, and sorry if it exists and I can't figure out a way to do it, but to have hi-res meshes exist only to cast "shadows" in the AO computation.
[/ QUOTE ]
Ok, for the next version perhaps. I need to close this beta bugs first.
Well... actually detected the bug... lol... I forgot to include the hxGridUserDLL.dll in the winstaller... hahah
Gonna post the final version in a few days tho
[/ QUOTE ]
Sweet, I grabbed the zip version and dropped the missing file into the folder and BINGO! 50 gHz of CPU action I can't wait to play with it now!
Sweet, I grabbed the zip version and dropped the missing file into the folder and BINGO! 50 gHz of CPU action I can't wait to play with it now!
[/ QUOTE ]
I love hxGrid because everybody can use the agent without using resources. It only uses CPU power when the computer is idle. We have installed it here in all the office's machines. Also is good for dedicated computation clusters.
I hope Roman could port it asap to win2000 and vista and x64.
The 3.11.0 final is almost ready. I'm revising all the AO alrogithms to get more quality and also waiting for Roman to solve some things.
Yeah the "Tab" thing for switching fields would be awesome, I've felt that missing a couple of times when trying to set up dimensions. So used to doing it in every other program, hopefully it's not difficult to set up
[ QUOTE ]
It won't let me generate non-square textures. What happens if I try is either it allocates the dimensions I give, but only uses a portion of the map to in effect render out a square map, or it crashes when starting the normal map generation.
[/ QUOTE ]
Ouch, you're right, gonna solve that for the final version! In the 3.11.0 beta 1 if you set width and height to a different value it explodes, indeed!
[ QUOTE ]
Oh another thing, why is there a warning to create a file that does *not* already exist?
[/ QUOTE ]
Damm you're a good beta tester I forgot to disable a flag for that dialog
[ QUOTE ]
There seems to be many unnecessary things like that in the program, like sounds, animations, illustrations, fancy programmer-art UI and so on.
[/ QUOTE ]
Yep! I used xNormal to experiment some thing I needed for other projects. For xNormal 4.0 I wanna use a more serious look. My idea is to avoid .NET(and any Microsoft-exclusive thing) at any cost and allow total user customization and total portability to other systems.
[ QUOTE ]
Oh and a request This may seem insignificant, but could it be possible to tab from the width to the heigh field in the "normal map" section? It would mean not having to use the mouse twice every time we want to set the map dimensions.
[/ QUOTE ]
Yep yep, I need to revise all the tab order in the dialogs.
more important suggestion: being able to manually edit the "output file" field. I instinctively try to do this a lot. Most of the time just to add numerals at the end of the filename for different versions of output.
[/ QUOTE ]
Due to the plugin nature of xNormal the user needs to select an input/ouput file + the plugin to use. Two different plugins can be assigned to the same file extension ( for example, Ogre .mesh and Stooge .mesh ). On the other hand two plugins can open the same file type(for example superPNG and the default PNG exporter... but superPNG compress more and better).
I could try to select a proper importer for the file, but that can generate potential conflicts(for example, I try to load a Stooge .mesh when is a Ogre one... Or I select the default PNG exporter but the user really wanted the superPNG one).
That's why I did the text comboboxes read only, to force the user to select the importer to use with the open/save dialog.
[ QUOTE ]
Maybe an automatic naming, even, would be good. I think for a good deal of us, most of the time the normal map will have the name of the lowpoly model + "_normal" or something. A good naming convention will even let viewers and such loads all maps automatically.
the lowpoly file is:
8ml_civilwar_union_soldier_body.mesh
which means the following must be the names for the maps generated, in the same path as the lowpoly. Always.
8ml_civilwar_union_soldier_body_norm.tga
8ml_civilwar_union_soldier_body_diff.tga
8ml_civilwar_union_soldier_body_ambocc.tga
[/ QUOTE ]
Is ok, but what happens when you use the batch rendering system? The main idea is to combine multiple meshes into the same normal map output. If I render a different map for each mesh then the batch rendering system benefits dissapear...
Also that file termination is not completely descriptive. The _norm is really normalsAndDisplacement(because the .a is the height and I cannot sepparate these due to SM2.0 limits). The ambocc is really occAndBentNormals. And the diffuse is not really "diffuse".. is a decal/albedo base texture.
[/ QUOTE ]
Oh and "background" color for normal maps should be 128,128,255 To make mipmapping totally safe and also tends to make it easier to copypaste in parts from other maps etc
[/ QUOTE ]
Omg, you found a board bug too It doesn't appear the "quote" above!
Okie. That's good too to mitigate seams.
Just a question... Want that too for object space ones?
[ QUOTE ]
Jogshy: About identical file extensions: Don't you check the file header anyway to avoid loading garbage? And if you do, then users shouldn't have to specify formats/plugins.
[/ QUOTE ]
Sometimes yes, sometimes nop. I use a lot of 3rd party libraries. Most of them aren't very well designed. Most don't throw exceptions or just return FALSE when they can't load a model because have no header info... they simply crash... I could filter these cases, but the real problem are the plugins which can read/write the same file type(like the superPNG and the PNG or the FBX/dotXSI+collada).
Well, let me see if I can do something with that.
[ QUOTE ]
What fools name their meshes .mesh anyway?
[/ QUOTE ]
Apparently a lot lol...Yepo, that's a bad thing haha
[ QUOTE ]
I would like to know if people here actually use several lowpoly models for one map, though, and why they can't just export those models into one file..
[/ QUOTE ]
I think people uses that to avoid "floating geometry" problems. Also for self occlusion problems in the AO and to combine normal maps. I think they will kill me if I turn off the batch rendering
[ QUOTE ]
Any chance of a rendering preview so you won't have to wait until the very end to find out something is wrong?
[/ QUOTE ]
Nope without increasing the rendering time. I could paint each triangle being rendered, but that will consume between 15 and 25% more time, .NET bitmap lockouts are very expensive.
I could preview it in realtime using D3D or OpenGL, but that will require a 3D card to generate them and I don't like the idea of increasing more the requisites to run the program.... I don't wanna require a 3D card(need SM2.0 as min I bet due to image normalization+HDR, SM1.0 is not enough) just for previewing in realtime the results heheh
so, if you wanna know if something is wrong the only solution is to render a 128x128 map to wait less and see
I'm preparing a GPU renderer and perhaps you could see the results in realtime, i'm not sure, is still not finished.
Is there a way to resize the windowed viewer? I am trying to capture a rotation video through camtasia. It does not see to like full screen 3dview. When I activate camtasia with F9 it closes the viewer. The windowed view works great but is not as large as I would like.
[ QUOTE ]
Is there a way to resize the windowed viewer?
Thanks
[/ QUOTE ]
Yep, but need to use the OpenGL graphics driver ( the D3D cannot resize window atm ).
So, change the default graphics driver to OpenGL going to the Plugin Manager.
Then, open the 3d viewer and put the cursor near the window border, click over and resize it. I corrected a bug about that recently, not sure if that works in the 3.11.0 beta 1 though.
I have a really high poly model that I created in Zbrush that Xnormal runs out of memeory trying to handle at normal render time. There has been mention of converting these files to a format that is better for this. From what I understand I need to open the viewer to then re-export my hi poly mesh in a different format, but it is too large to open in the viewer. Any other options?
There has been mention of converting these files to a format that is better for this. From what I understand I need to open the viewer to then re-export my hi poly mesh in a different format, but it is too large to open in the viewer. Any other options?
[/ QUOTE ]
Just need to open in the viewer and re-export to modify the lowpoly cage information.
If you highpoly model is too large you can try to split the object into different parts and render them sepparately. If the OBJ is taking too much time to load try to save a SBM with 3dsmax, is the fastest format(the LWO and FBX are fast too).
[ QUOTE ]
I have a really high poly model that I created in Zbrush that Xnormal runs out of memeory trying to handle at normal render time.
[/ QUOTE ]
Tomorrow I will upload the final version. It includes a new memory manager which should anihilate the previous one
- Added full screen antialiasing support for the DX9 graphics driver.
- Solved problems with ALT+TAB and also with the mouse.
- Added a checkbox inside the 3D viewer to lock the position of the light to the camera. Notice the shadows dissapear when you enable this mode(which is logical on the other hand)...
- ATI viewer problems solved(I hope...)
- Changed the default normal map back color to that nice blue we all luv!
- Added a checkbox to auto center the camera orbit point in the 3D viewer.
- Distributed rendering with hxGrid. Notice it doesn't work yet with Win2000, but should work with XP SP1 or above and Vista.
- Added "attenuation" and "jitter" options to the ambient occlusion generation. Also improved a lot the quality of the AO generation. Caution when you set the # of rays... now with 32 you can get the same results than 200 before.
- Modified internally the memory manager to allow to use RAM more efficiently. That increases a lot the polygon limits.
- Solved a crash bug when you set the render maps width and height different.
- Added a two new Photoshop filters: the dilation filter and the normalize filter.
- Now the grid is a plane, not lines.
- Solved a problem that was causing the 3DSMAX6/7/8 exporter not to be loaded using 3DSMAX 6 and 7.
- Added projective caustics/translucent shadows to the 3D viewer ( see Bojana's example ).
- Now the DX9 graphics driver will use high-quality floating-point normal maps if your graphics card can filter/mipmap them correctly.
- Added axis indicators to the referenceCubemap.jpg.
- Rolled back to the DirectX April 2007 edition due to problems with the June 2007 one ( debug mode is broken, speed loss, etc... ).
- Added a new tool: the object/tangent space converter.
- Added a "mesh scale" parameter to control the conversion units when importing geometry.
- Solved some hang problems with multi-core CPUs when showing the progress window.
- Corrected tons of bugs and improved the UI accesibility with tab order.
Enjoy it!
ps: I'm working on faster map renderers for the next version... but will require dedicated hardware.
Do I need to rollback DX? I am having issues getting any AO maps to render. It just sits there saying Rendering Occlusion map.. No updating or anything. I have it set to 20 rays. The normal map rendered great.
I am having issues getting any AO maps to render. It just sits there saying Rendering Occlusion map.. No updating or anything. I have it set to 20 rays. The normal map rendered great.
[/ QUOTE ]
Hmmmm curious. Got a problem with that but I think solved that bug... Are you using the x64 version and/or Vista? Do you have a dual core CPU? Well, if the x64 version gives you problems try with the x86 which has been much more tested.
Does the smiley example work if you turn on the AO generation? 16-20 rays is fine to test. Caution... sure you set rays from 16 to 32 to test... the old xNormal versions used typically 50-100 rays and the settings are saved when you close the program.. so perhaps is using 100 rays settings from old versions and 100 rays for the new one is a lot!
Sometimes happens that if your model is not enough tessellated the progress bar updates suddenly
AO can be a bit slow, I know... i'm working in a new renderer to accelerate it in the faogen line.
i've had the same problem with the ao map generation. was able to do a 512 map without a prob, but tried 2048 and it froze on me twice.
using x86 on single core intel proc, win xp, default scanline renderer.
edit: forgot to mention--love this program!!! thanks so much!
Hmmm I just notice a thing importing objects from ZB3.... I got all the face normals flipped by default... looks like I must invert them... but I did the smiley with ZB2 and worked ok... does ZB3 changed the face vertex order? Any idea? I think need to put an option to flip them or something...
[ QUOTE ]
If I alt-tab out and comeback I can't select anything;
[/ QUOTE ]
Remember to press ALT+ENTER to release/catch the mouse.
[ QUOTE ]
I'm lovin the camera auto center and lock light to camera, superlative work!!
[/ QUOTE ]
Btw I detected a bug with the camera... sometimes locks ... if that happens please press the autofit scene button.
Gonna solve it for the next version.
I notice you can import HDR images in the engine. When I add them to reflection I don't really see any difference betweem them and a regular jpg. Is this currently supported?
[ QUOTE ]
I notice you can import HDR images in the engine. When I add them to reflection I don't really see any difference betweem them and a regular jpg. Is this currently supported?
[/ QUOTE ]
The 3d viewer does not use high dynamic range at the moment. That's because some graphics cards don't support floating point filtering / mipmapping. The DX10/LongsPeak driver i'm preparing will support it ( as well as other advanced features like micro triangle displacement ).
[ QUOTE ]
Blender got a lot of exporters but can you please make Xnormal support the .blend format?
[/ QUOTE ]
I was trying to support Ogre3D and Blender for the next version... But i'm not sure on Blender... I think they use a super-slow Python thing which gonna be a pain to import/export a 4M poly mesh.
Having lots of issues with the model viewer portion with the latest available version.
Locks up when trying to enable/disable anything. Also performance is quite poor. 3.11 beta 1 performance was around 60 fps with arsh's diablo and now i'm lucky to get 30fps in a window. My normal res is 1600x1200. I also noticed I was only given the option of my desktop res with 4x, 2x or no AA.
System specs:
p4 2.0ghz
1024mb ram
geforce 6600gt
vista
I am probably missing some details, but the basics is that the viewer just fails massively now for me.
[ QUOTE ]
Locks up when trying to enable/disable anything. Also performance is quite poor. 3.11 beta 1 performance was around 60 fps with arsh's diablo and now i'm lucky to get 30fps in a window. My normal res is 1600x1200. I also noticed I was only given the option of my desktop res with 4x, 2x or no AA.
[/ QUOTE ]
Try to disable the shadows, the emisive glow and the AA... are very expensive. You can also try to use the OpenGL driver instead the DX9 one ( in the plugin manager ).
Due to the strange nature of Direct3D 9 if you ALT+TAB in fullscreen mode the device will reset. When this happens I need to reload all the textures, meshes, etc... unfortunally artists have the bad habit to ALT+TAB a lot to swap applications to retouch textures, etc... Most of the DX9 games just crash if you do an ALT+TAB in fullscreen mode.
Due to that, I decided not to use a real fullcreen mode... the 3.11.0 final just runs a big windowed mode with a window without border/title bar. That can hurt a bit the performance but you gain some benifits, like full ALT+TAB support, solves dialog problems, improves multimonitor compatibility, allows to take screenshot more easy, etc... That's why the video modes are locked to the desktop ones... because are not real full screen modes.
Perhaps I need to rollback to the old system and solve the ALT+TAB in other way... Yep, need to think about it.
[ QUOTE ]
System specs:
p4 2.0ghz
1024mb ram
geforce 6600gt
vista
[/ QUOTE ]
Vista drivers are a pain at the moment. Nothing is well optimized, OpenGL crashes a lot and the x64 system is pretty unstable, consumes tons of RAM and resources, tons of programs crash or aren't Vista-ready yet, etc... Also have some serious concept errors like the multimonitor limitations(cards need to share the same driver), the UAC and folder rights problems, the digital signing warnings, DRM, mobile battery eat using Aero, excesive minimum requisites, etc...
I support Vista because I have to but I don't like that OS at all. Vista = pain
I got that error emailed some time ago but couldn't determine how to solve it. Is a rare problem in the GDI+ component of the .NET 2.0. Gonna try to take an alternative path for the next version but i'm not sure if I could fix it.... I tryed with a GF6800 + Vista x64 + FW162.22 and couldn't repro Well, gonna see what I can do.
[ QUOTE ]
1) It would be nice to be able to give every model a color (no texture), which would be baked as a color map.
[/ QUOTE ]
Like an objectID? Well I can output an objectID map very easy if ya want. Or are your referring to give a random color to the non-texture meshes inside the 3d viewer?
[ QUOTE ]
2) Programms like RoadKill have a Maya Plugin:
You select your models and Press a Button (in maya) and xnormal opens with all the models loaded.
[/ QUOTE ]
I think XSI has that too ( Crosswalk? )
Ok, gonna investigate that but my Maya SDK knowledge needs some improvement heheh! I'm sure is possible to transfer max, maya and xsi data directly to xnormal without exporting any model... in fact asset managers like Alienbrain can do that putting a small window menu.. but no idea how they do it
ps. The 3.11.1 is almost ready with two new HW-GPU-accelerated renderers and some bugs corrected.
Ok, gonna implement that.. but the question is... wanna really select manually the colors or can I use pseudo-random objectID color better?
The objectID could work as follows... If you don't specify a base texture file then will assign a pseudo-random color to the object, so all the pixels those belong to it will be painted in the base texture color map with that color. Is that ok?
Hmmmmm... I just noticed a thing....
A question... Does Photoshop CS3 x64 bits exist or is a x86 program certified to run in a x64 operating system? I just noticed the Photoshop filters don't work under Vista x64... because the Photoshop installer installs Photoshop CS3 x86 in the Program files(x86)\Adobe\Photoshop CS3 .... so Photoshop CS3 is a 32bits application always but I compiled some filters for x64. Did not Abobe release a full x64 binary for PSCS3 omg?
This is probably a really stupid question, but I've been trying to normal map a knife. The high poly is exported from zbrush3, and the low from 3dsm. Every time I render, I'm not getting the uv layout from the lowpoly, but instead it's just a planar map. I've tried various formats for the lowpoly, but I can't really edit the high because my system can't handle it.
I'm not sure what you mean, I only used 1 channel when I unwrapped it, it will eventually go into Source. When I import the .3ds into max, I can still texport the maps perfectly, so I don't think it's a problem with the formats or anything.
i give up! i cant render AO maps with XNORMAL!
is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar. i mean. the green light still lights up but come on!!
my hipoly model is like 1mill polys.
I can't even install xNormal. It quits after the unpack telling me to install DirectX 9.0c April 2007 or above and redirects me to microsoft's website. I updated DirectX...numerous times. It still gives me that error.
dxdiag even shows I have version 9.0c 4.09, plus I have the original xNormal installed and running
Ahh, it seems I exported my reference plane along with my lowpoly, which was also scaled incorrectly.
[ QUOTE ]
i give up! i cant render AO maps with XNORMAL!
is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar. i mean. the green light still lights up but come on!!
my hipoly model is like 1mill polys.
Is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar.
[/ QUOTE ]
I think there is a bug with some dual core machines. It is curious but if you render AO again without closing the program it takes much more time to render, is strange.
Do you see any difference running the hxGrid renderer with 1/2 machines instead of the default one?
The progress bar just updates each lowpoly.nFaces/20 steps... so if your model is not very tessellated won't notify percentage too much. Windows multitaskting model don't help too.
I'm preparing a HW-accelerated AO renderer to help
[ QUOTE ]
It quits after the unpack telling me to install DirectX 9.0c April 2007 or above and redirects me to microsoft's website. I updated DirectX...numerous times. It still gives me that error.
dxdiag even shows I have version 9.0c 4.09, plus I have the original xNormal installed and running
[/ QUOTE ]
The dx 9.0c 4.09 just indicates you have DX9.0c installed, not the concrete revision.
To know if you have installed the correct DX9 April 2007 "revision" need to see if the d3dx9_33.dll exists on the system directory(usually c:\windows\system32 on winXP )
If you are 100% sure you have the correct DX version installed(and the .NET 2.0 + VS2005SP1 DLLs) you can try to use the 7zipped version (which occupies less)... if you see any "sorry can't find d3dx9_XXXXX.dll while the program loads the plugins then you don't have the April 2007 installed.
I'm trying to release the 3.11.1 in a few, but hey! i'm on vacations lost in the middle of an isle without a computer near until September, I wrote this using my Dune telephaty!
Replies
In the ambient occlusion options, what is the Bias value for? I don't see an explanation for it in the help file.
[/ QUOTE ]
It basically extrudes a bit the highpoly hit point to solve some self-occlusion problems. It works in the same way that Blender... see this image:
Bias 0.0001:
Notice the self occlusion problem at the triangle edges...
If you increment the AO bias a few to 0.015:
Also, playing with the AO angle aperture degrees you can get a dirtier AO map or cleaner(or touch the AO method for a more-drastic result.. from cosine to uniform, for example)
And remember... you can(optionally) use the cage/maximum uniform front/back ray distance to control the ray extension longitude... or uncheck the "don't limit ray distance" to disable it.
Also remember you don't need a highpoly+lowpoly model to render the AO. You can render the AO with just one model(selecting it as lowpoly and highpoly, the same file in the slots). If you use this, sure you set well the cage+a greater AO bias or you will get a void AO map!(because rays won't collide themselves).
You could use too the new render fast AO with model+normal map that Jonhy luvs!(hope)
ps: Omg Fat, I saw that nice hair video in your web and now I can't stop thinking about how to implement it in realtime! Why why why you put there!!! Now I can't sleep!
And I noticed that a couple people here have said that the ambient occlusion from xNormal isn't as good as other packages, so I did some tests. For some reason you don't seem to get the detail in xNormal that you do in Max, for example. Here's a picture...
I tried to get the settings between the Mental Ray AO shader and xNormal's AO as close as possible. But you can see that there's more detail in the MR rip. I couldn't say why, but I think that's what people are talking about. Any thoughts on where the difference is coming from?
You mean the hair on the guy? I did that with HairFX in Max. If you can get that in real-time, you'll have a lot of people knocking on your door.
[/ QUOTE ]
Yep, was that
[ QUOTE ]
For some reason you don't seem to get the detail in xNormal that you do in Max, for example. Here's a picture...
[/ QUOTE ]
Looks like you need to play a bit with the Ao angle(tru to increase or decrease it). Play too a bit with min/max output levels, the right image looks a bit saturated.
Btw, caution if you are saving to OpenEXR file format ( need to set the exposure and gamma well when importing/vieweing the file in Photoshop, etc ) or the white color will "saturate"
Also try to reduce the AO bias(but I think the default is the minimum...). Try always to put it to zero, unless artifacts appear(like in the Blender ball I posted before).
Also very important, set the minimum #of rays and maxiumum EQUAL ( so adaptative AO will be disabled ).
Also sure don't use the black sand median filter neither the blur filter.
Finally, try to scale the mesh up. To avoid small floating point errors I extrude a constant distance(something like 0.000001f) the ao rays origin so scaling the mesh a bit up should help.
But well... to get the desired microdetail there probably you should use a cavity map(or a thing I have here prepared for the next version
And ... MR uses area-51 technology(deepshadow volexized rays,energy radiance dispersion with BDRF,128bits floating point precision, micro-patches ala REYES,blah blah) I can't barely dream
Well, let me revise a few the code if I can find something wrong. Actually I think I know what's happening( I made a mistake in the AO formulae, but i'm not completely sure... let me make some tests).
ps: I don't abuse the
Feature Idea - It would be nice, and sorry if it exists and I can't figure out a way to do it, but to have hi-res meshes exist only to cast "shadows" in the AO computation. These meshes are not calculated in any other way, they don't get hit with normal rays, they don't get hit with Ao rays, etc.
Why would one need this? Example. I have a character with a great deal of armor collar around his neck. This collar needs to be a separate model. I now want to calcualate the head and neck AO but the collar is so close there is no way for me to not get eithe rmy cage or rays to not intersect it in some way. Also if I use rays it is much easier to process the head on it's own pass. Now if I could load the collar up to "Cast AO ONLY" it would be almost as if it did not exist in the calculations except to cast shadows on the hi poly head that are then calcualted in the heads AO map. I hope that makes some sense?
Thanks
Jesse
Feature Idea - It would be nice, and sorry if it exists and I can't figure out a way to do it, but to have hi-res meshes exist only to cast "shadows" in the AO computation.
[/ QUOTE ]
Ok, for the next version perhaps. I need to close this beta bugs first.
Well... actually detected the bug... lol... I forgot to include the hxGridUserDLL.dll in the winstaller... hahah
Gonna post the final version in a few days tho
[/ QUOTE ]
Sweet, I grabbed the zip version and dropped the missing file into the folder and BINGO! 50 gHz of CPU action
Sweet, I grabbed the zip version and dropped the missing file into the folder and BINGO! 50 gHz of CPU action
[/ QUOTE ]
I love hxGrid because everybody can use the agent without using resources. It only uses CPU power when the computer is idle. We have installed it here in all the office's machines. Also is good for dedicated computation clusters.
I hope Roman could port it asap to win2000 and vista and x64.
The 3.11.0 final is almost ready. I'm revising all the AO alrogithms to get more quality and also waiting for Roman to solve some things.
It won't let me generate non-square textures. What happens if I try is either it allocates the dimensions I give, but only uses a portion of the map to in effect render out a square map, or it crashes when starting the normal map generation.
[/ QUOTE ]
Ouch, you're right, gonna solve that for the final version! In the 3.11.0 beta 1 if you set width and height to a different value it explodes, indeed!
[ QUOTE ]
Oh another thing, why is there a warning to create a file that does *not* already exist?
[/ QUOTE ]
Damm you're a good beta tester
[ QUOTE ]
There seems to be many unnecessary things like that in the program, like sounds, animations, illustrations, fancy programmer-art UI and so on.
[/ QUOTE ]
Yep! I used xNormal to experiment some thing I needed for other projects. For xNormal 4.0 I wanna use a more serious look. My idea is to avoid .NET(and any Microsoft-exclusive thing) at any cost and allow total user customization and total portability to other systems.
[ QUOTE ]
Oh and a request
[/ QUOTE ]
Yep yep, I need to revise all the tab order in the dialogs.
thx for all the comments!
more important suggestion: being able to manually edit the "output file" field. I instinctively try to do this a lot. Most of the time just to add numerals at the end of the filename for different versions of output.
[/ QUOTE ]
Due to the plugin nature of xNormal the user needs to select an input/ouput file + the plugin to use. Two different plugins can be assigned to the same file extension ( for example, Ogre .mesh and Stooge .mesh ). On the other hand two plugins can open the same file type(for example superPNG and the default PNG exporter... but superPNG compress more and better).
I could try to select a proper importer for the file, but that can generate potential conflicts(for example, I try to load a Stooge .mesh when is a Ogre one... Or I select the default PNG exporter but the user really wanted the superPNG one).
That's why I did the text comboboxes read only, to force the user to select the importer to use with the open/save dialog.
[ QUOTE ]
Maybe an automatic naming, even, would be good. I think for a good deal of us, most of the time the normal map will have the name of the lowpoly model + "_normal" or something. A good naming convention will even let viewers and such loads all maps automatically.
the lowpoly file is:
8ml_civilwar_union_soldier_body.mesh
which means the following must be the names for the maps generated, in the same path as the lowpoly. Always.
8ml_civilwar_union_soldier_body_norm.tga
8ml_civilwar_union_soldier_body_diff.tga
8ml_civilwar_union_soldier_body_ambocc.tga
[/ QUOTE ]
Is ok, but what happens when you use the batch rendering system? The main idea is to combine multiple meshes into the same normal map output. If I render a different map for each mesh then the batch rendering system benefits dissapear...
Also that file termination is not completely descriptive. The _norm is really normalsAndDisplacement(because the .a is the height and I cannot sepparate these due to SM2.0 limits). The ambocc is really occAndBentNormals. And the diffuse is not really "diffuse".. is a decal/albedo base texture.
[/ QUOTE ]
Oh and "background" color for normal maps should be 128,128,255
[/ QUOTE ]
Omg, you found a board bug too
Okie. That's good too to mitigate seams.
Just a question... Want that too for object space ones?
Jogshy: About identical file extensions: Don't you check the file header anyway to avoid loading garbage? And if you do, then users shouldn't have to specify formats/plugins.
[/ QUOTE ]
Sometimes yes, sometimes nop. I use a lot of 3rd party libraries. Most of them aren't very well designed. Most don't throw exceptions or just return FALSE when they can't load a model because have no header info... they simply crash... I could filter these cases, but the real problem are the plugins which can read/write the same file type(like the superPNG and the PNG or the FBX/dotXSI+collada).
Well, let me see if I can do something with that.
[ QUOTE ]
What fools name their meshes .mesh anyway?
[/ QUOTE ]
Apparently a lot lol...Yepo, that's a bad thing haha
[ QUOTE ]
I would like to know if people here actually use several lowpoly models for one map, though, and why they can't just export those models into one file..
[/ QUOTE ]
I think people uses that to avoid "floating geometry" problems. Also for self occlusion problems in the AO and to combine normal maps. I think they will kill me if I turn off the batch rendering
[ QUOTE ]
Any chance of a rendering preview so you won't have to wait until the very end to find out something is wrong?
[/ QUOTE ]
Nope without increasing the rendering time. I could paint each triangle being rendered, but that will consume between 15 and 25% more time, .NET bitmap lockouts are very expensive.
I could preview it in realtime using D3D or OpenGL, but that will require a 3D card to generate them and I don't like the idea of increasing more the requisites to run the program.... I don't wanna require a 3D card(need SM2.0 as min I bet due to image normalization+HDR, SM1.0 is not enough) just for previewing in realtime the results heheh
so, if you wanna know if something is wrong the only solution is to render a 128x128 map to wait less and see
I'm preparing a GPU renderer and perhaps you could see the results in realtime, i'm not sure, is still not finished.
[/ QUOTE ]
thx for the feedback!
Thanks
Is there a way to resize the windowed viewer?
Thanks
[/ QUOTE ]
Yep, but need to use the OpenGL graphics driver ( the D3D cannot resize window atm ).
So, change the default graphics driver to OpenGL going to the Plugin Manager.
Then, open the 3d viewer and put the cursor near the window border, click over and resize it. I corrected a bug about that recently, not sure if that works in the 3.11.0 beta 1 though.
I have a really high poly model that I created in Zbrush that Xnormal runs out of memeory trying to handle at normal render time. There has been mention of converting these files to a format that is better for this. From what I understand I need to open the viewer to then re-export my hi poly mesh in a different format, but it is too large to open in the viewer. Any other options?
Thanks yet again
Jesse
There has been mention of converting these files to a format that is better for this. From what I understand I need to open the viewer to then re-export my hi poly mesh in a different format, but it is too large to open in the viewer. Any other options?
[/ QUOTE ]
Just need to open in the viewer and re-export to modify the lowpoly cage information.
If you highpoly model is too large you can try to split the object into different parts and render them sepparately. If the OBJ is taking too much time to load try to save a SBM with 3dsmax, is the fastest format(the LWO and FBX are fast too).
[ QUOTE ]
I have a really high poly model that I created in Zbrush that Xnormal runs out of memeory trying to handle at normal render time.
[/ QUOTE ]
Tomorrow I will upload the final version. It includes a new memory manager which should anihilate the previous one
- Added full screen antialiasing support for the DX9 graphics driver.
- Solved problems with ALT+TAB and also with the mouse.
- Added a checkbox inside the 3D viewer to lock the position of the light to the camera. Notice the shadows dissapear when you enable this mode(which is logical on the other hand)...
- ATI viewer problems solved(I hope...)
- Changed the default normal map back color to that nice blue we all luv!
- Added a checkbox to auto center the camera orbit point in the 3D viewer.
- Distributed rendering with hxGrid. Notice it doesn't work yet with Win2000, but should work with XP SP1 or above and Vista.
- Added "attenuation" and "jitter" options to the ambient occlusion generation. Also improved a lot the quality of the AO generation. Caution when you set the # of rays... now with 32 you can get the same results than 200 before.
- Modified internally the memory manager to allow to use RAM more efficiently. That increases a lot the polygon limits.
- Solved a crash bug when you set the render maps width and height different.
- Added a two new Photoshop filters: the dilation filter and the normalize filter.
- Now the grid is a plane, not lines.
- Solved a problem that was causing the 3DSMAX6/7/8 exporter not to be loaded using 3DSMAX 6 and 7.
- Added projective caustics/translucent shadows to the 3D viewer ( see Bojana's example ).
- Now the DX9 graphics driver will use high-quality floating-point normal maps if your graphics card can filter/mipmap them correctly.
- Added axis indicators to the referenceCubemap.jpg.
- Rolled back to the DirectX April 2007 edition due to problems with the June 2007 one ( debug mode is broken, speed loss, etc... ).
- Added a new tool: the object/tangent space converter.
- Added a "mesh scale" parameter to control the conversion units when importing geometry.
- Solved some hang problems with multi-core CPUs when showing the progress window.
- Corrected tons of bugs and improved the UI accesibility with tab order.
Enjoy it!
ps: I'm working on faster map renderers for the next version... but will require dedicated hardware.
I hope I have time to give it a spin today at work. Thanks for all your hard work on this. It is a great tool.
Do I need to rollback DX?
[/ QUOTE ]
Nope
[ QUOTE ]
I am having issues getting any AO maps to render. It just sits there saying Rendering Occlusion map.. No updating or anything. I have it set to 20 rays. The normal map rendered great.
[/ QUOTE ]
Hmmmm curious. Got a problem with that but I think solved that bug... Are you using the x64 version and/or Vista? Do you have a dual core CPU? Well, if the x64 version gives you problems try with the x86 which has been much more tested.
Does the smiley example work if you turn on the AO generation? 16-20 rays is fine to test. Caution... sure you set rays from 16 to 32 to test... the old xNormal versions used typically 50-100 rays and the settings are saved when you close the program.. so perhaps is using 100 rays settings from old versions and 100 rays for the new one is a lot!
Sometimes happens that if your model is not enough tessellated the progress bar updates suddenly
AO can be a bit slow, I know... i'm working in a new renderer to accelerate it in the faogen line.
using x86 on single core intel proc, win xp, default scanline renderer.
edit: forgot to mention--love this program!!! thanks so much!
I found you need to have iFlipX, iFlipY, eFlipX, eFlipY to get proper import/export results for Z3
Preferences/Importerexporter
I found you need to have iFlipX, iFlipY, eFlipX, eFlipY to get proper import/export results for Z3
[/ QUOTE ]
Nice thx!
Thanks Jogshy!
Just noticed the alt+tab problem still persists in the 3d viewer. If I alt-tab out and comeback I can't select anything;
I'm lovin the camera auto center and lock light to camera, superlative work!!
If I alt-tab out and comeback I can't select anything;
[/ QUOTE ]
Remember to press ALT+ENTER to release/catch the mouse.
[ QUOTE ]
I'm lovin the camera auto center and lock light to camera, superlative work!!
[/ QUOTE ]
Btw I detected a bug with the camera... sometimes locks ... if that happens please press the autofit scene button.
Gonna solve it for the next version.
I notice you can import HDR images in the engine. When I add them to reflection I don't really see any difference betweem them and a regular jpg. Is this currently supported?
[/ QUOTE ]
The 3d viewer does not use high dynamic range at the moment. That's because some graphics cards don't support floating point filtering / mipmapping. The DX10/LongsPeak driver i'm preparing will support it ( as well as other advanced features like micro triangle displacement ).
Blender got a lot of exporters but can you please make Xnormal support the .blend format?
[/ QUOTE ]
I was trying to support Ogre3D and Blender for the next version... But i'm not sure on Blender... I think they use a super-slow Python thing which gonna be a pain to import/export a 4M poly mesh.
Locks up when trying to enable/disable anything. Also performance is quite poor. 3.11 beta 1 performance was around 60 fps with arsh's diablo and now i'm lucky to get 30fps in a window. My normal res is 1600x1200. I also noticed I was only given the option of my desktop res with 4x, 2x or no AA.
System specs:
p4 2.0ghz
1024mb ram
geforce 6600gt
vista
I am probably missing some details, but the basics is that the viewer just fails massively now for me.
Locks up when trying to enable/disable anything. Also performance is quite poor. 3.11 beta 1 performance was around 60 fps with arsh's diablo and now i'm lucky to get 30fps in a window. My normal res is 1600x1200. I also noticed I was only given the option of my desktop res with 4x, 2x or no AA.
[/ QUOTE ]
Try to disable the shadows, the emisive glow and the AA... are very expensive. You can also try to use the OpenGL driver instead the DX9 one ( in the plugin manager ).
Due to the strange nature of Direct3D 9 if you ALT+TAB in fullscreen mode the device will reset. When this happens I need to reload all the textures, meshes, etc... unfortunally artists have the bad habit to ALT+TAB a lot to swap applications to retouch textures, etc... Most of the DX9 games just crash if you do an ALT+TAB in fullscreen mode.
Due to that, I decided not to use a real fullcreen mode... the 3.11.0 final just runs a big windowed mode with a window without border/title bar. That can hurt a bit the performance but you gain some benifits, like full ALT+TAB support, solves dialog problems, improves multimonitor compatibility, allows to take screenshot more easy, etc... That's why the video modes are locked to the desktop ones... because are not real full screen modes.
Perhaps I need to rollback to the old system and solve the ALT+TAB in other way... Yep, need to think about it.
[ QUOTE ]
System specs:
p4 2.0ghz
1024mb ram
geforce 6600gt
vista
[/ QUOTE ]
Vista drivers are a pain at the moment. Nothing is well optimized, OpenGL crashes a lot and the x64 system is pretty unstable, consumes tons of RAM and resources, tons of programs crash or aren't Vista-ready yet, etc... Also have some serious concept errors like the multimonitor limitations(cards need to share the same driver), the UAC and folder rights problems, the digital signing warnings, DRM, mobile battery eat using Aero, excesive minimum requisites, etc...
I support Vista because I have to but I don't like that OS at all. Vista = pain
noticed another problem at least within vista where the gui is broken slightly. Everything works except things being offset.
noticed another problem at least within vista where the gui is broken slightly. Everything works except things being offset.
[/ QUOTE ]
Usually that happens if:
1) You have more than 3Gb RAM without the /3G /userva option in the boot.ini
2) Defective video drivers ( which ForceWare are you using? )
3) The damm http://weblogs.asp.net/okloeten/archive/2007/07/11/3103882.aspx bug. Microsoft never patched that and has no intention to solve it.
I got that error emailed some time ago but couldn't determine how to solve it. Is a rare problem in the GDI+ component of the .NET 2.0. Gonna try to take an alternative path for the next version but i'm not sure if I could fix it.... I tryed with a GF6800 + Vista x64 + FW162.22 and couldn't repro
There are two thigns I miss though
1) It would be nice to be able to give every model a color (no texture), which would be baked as a color map.
2) Programms like RoadKill have a Maya Plugin:
You select your models and Press a Button (in maya) and xnormal opens with all the models loaded.
1) It would be nice to be able to give every model a color (no texture), which would be baked as a color map.
[/ QUOTE ]
Like an objectID? Well I can output an objectID map very easy if ya want. Or are your referring to give a random color to the non-texture meshes inside the 3d viewer?
[ QUOTE ]
2) Programms like RoadKill have a Maya Plugin:
You select your models and Press a Button (in maya) and xnormal opens with all the models loaded.
[/ QUOTE ]
I think XSI has that too ( Crosswalk? )
Ok, gonna investigate that but my Maya SDK knowledge needs some improvement heheh! I'm sure is possible to transfer max, maya and xsi data directly to xnormal without exporting any model... in fact asset managers like Alienbrain can do that putting a small window menu.. but no idea how they do it
ps. The 3.11.1 is almost ready with two new HW-GPU-accelerated renderers and some bugs corrected.
http://www.softimage.com/products/xsi/videos.aspx?video_id=10 <- video of it in action.
something like that...
getting things from maya to any program seems to be very easy with mel.
Just check RoadKill.mel
The objectID could work as follows... If you don't specify a base texture file then will assign a pseudo-random color to the object, so all the pixels those belong to it will be painted in the base texture color map with that color. Is that ok?
looking foreward to this feature
A question... Does Photoshop CS3 x64 bits exist or is a x86 program certified to run in a x64 operating system? I just noticed the Photoshop filters don't work under Vista x64... because the Photoshop installer installs Photoshop CS3 x86 in the Program files(x86)\Adobe\Photoshop CS3 .... so Photoshop CS3 is a 32bits application always but I compiled some filters for x64. Did not Abobe release a full x64 binary for PSCS3 omg?
is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar. i mean. the green light still lights up but come on!!
my hipoly model is like 1mill polys.
dxdiag even shows I have version 9.0c 4.09, plus I have the original xNormal installed and running
[ QUOTE ]
i give up! i cant render AO maps with XNORMAL!
is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar. i mean. the green light still lights up but come on!!
my hipoly model is like 1mill polys.
[/ QUOTE ]
same!
Is xnormal suposed to take ages to render the map? it looks like it stops at the middle of the progress bar.
[/ QUOTE ]
I think there is a bug with some dual core machines. It is curious but if you render AO again without closing the program it takes much more time to render, is strange.
Do you see any difference running the hxGrid renderer with 1/2 machines instead of the default one?
The progress bar just updates each lowpoly.nFaces/20 steps... so if your model is not very tessellated won't notify percentage too much. Windows multitaskting model don't help too.
I'm preparing a HW-accelerated AO renderer to help
[ QUOTE ]
It quits after the unpack telling me to install DirectX 9.0c April 2007 or above and redirects me to microsoft's website. I updated DirectX...numerous times. It still gives me that error.
dxdiag even shows I have version 9.0c 4.09, plus I have the original xNormal installed and running
[/ QUOTE ]
The dx 9.0c 4.09 just indicates you have DX9.0c installed, not the concrete revision.
To know if you have installed the correct DX9 April 2007 "revision" need to see if the d3dx9_33.dll exists on the system directory(usually c:\windows\system32 on winXP )
If you are 100% sure you have the correct DX version installed(and the .NET 2.0 + VS2005SP1 DLLs) you can try to use the 7zipped version (which occupies less)... if you see any "sorry can't find d3dx9_XXXXX.dll while the program loads the plugins then you don't have the April 2007 installed.
I'm trying to release the 3.11.1 in a few, but hey! i'm on vacations lost in the middle of an isle without a computer near until September, I wrote this using my Dune telephaty!