I changed a lot the CPU multicore code. Please tell me if you find any error/bug/crash!
The final version will include a lot of new things, but I need to be sure first that the new multicore renderer works ok! So that's why a beta was released.
- New render AO without highpoly feature with normal map support ( specially dedicated to Johny! ) Select your lowpoly model in both the highpoly/lowpoly grid, specify a normal map(optional) and you will get fast and cheap AO approximation ( and remember to use cages to limit distance for better results ). Also can be used to get more detail in the default AO system. What does is to invalidate the hit surface normal and use the normal present in the normal map.
- Two new tools: height map to occlusion map + PhotoNormal to make normal maps directly from photographs!
- Added two new tangent basis calculators(D3DX,MeshMender) to improve compatibility with programs and 3d engines.
- Added the 8MonkeyLabs mesh importer for their upcoming game Darkest of days ( I heard this game gonna revolutionise the FPS shooter genere! )
- Re-structured the program internally to be more memory efficient and also coded again the multicore/multithread routines( I fear that, pls let me know if you have any crash problem related to this ).
- Added camera/light automatic movement inside the 3D viewer
- Solved multiple bugs ( Photoshop filters displaced bitmap preview, grayscale JPG import error, OpenGL window resize, interior AO, incorrect external cage warning, etc... ) and also i'm pretty sure I added some new bugs too
Btw, requires the latest June 2007 DirectX, sorry I'm a M$ofer
ps: Already working in the 3.11.0 which gonna include a new and amazing fast renderer. But that will be after the summer, vacations are just here !
hey jogshy , looking polyshed as always, the normal map generates lightning fast !!!! and with AA in max ! the only issue i have is with the cavity mapping tool, became too confusing and takes along time to generate anything
[ QUOTE ]
the only issue i have is with the cavity mapping tool, became too confusing and takes along time to generate anything
[/ QUOTE ]
The EPD method is the fastest. For the Min,max,avg,spd try to reduce the search radius from 8 pixels to 3/4. (Min/Max are strange, use avg/spd usually ).
If you have a lowpoly mesh + a normal map better use the new render AO with normal map feature, probably will give you better results.
About the render speed I think i can improve it a lot ( specially the AO). In fact, the new 3.11 i'm preparing will smash all you've seen!
[ QUOTE ]
hey jogshy, another thing, the 3d viewer doesnt work for me, i mean it works, but i cant use the mouse, the cursor gets stuck in the upper corner
[/ QUOTE ]
Are you using the windowed mode or full screen? If you are using windowed mode remember to ALT+ENTER to release/capture the mouse cursor.
Also try to center the window, there is a bug currently with the mouse cursor if the window is clipped( where 'clipped' means you move a part the window outside the monitor). That will be patched shortly.
so one thing i've noticed, and this is kind of silly. but often times it seems that whats taking longest to render is the dilation filter(aside from ambocc of course) i'll have the normals calculated in a matter of seconds and it will take twice as long just to do the 8 passes of dilation on a large map like a 2048. Also sometimes it will go through this process 2 or 3 times. Is there any way you go just stretch the edges out in 1 pass, instead of doing a seperate pass for every level?
think about how dilation works, it cannot get accelerated much at all. For each pixel you must check if one of the neighbors (4 or 8) was set and then copy its value. In the 2k x 2k texture map thats like 33 million lookups. One could try to do it on the GPU, by using the stencil buffer and offset the image by one pixel, so 8 passes each reflection one of the directional shifts. But I am not sure if that is really faster than the CPU.
One thing to consider is that pixels of previous/next row are "far away" in the memory, which makes things slower.
btw awesome work jogshy, that you continue your perfect support for this tool for so long, is really top notch. I hope some of you pros using this tool frequently remember to "donate" something back... Yes it's for free, but if you use it to make money, you should not forget the immense workload people put in free software...
i'll have the normals calculated in a matter of seconds and it will take twice as long just to do the 8 passes of dilation on a large map like a 2048. Also sometimes it will go through this process 2 or 3 times. Is there any way you go just stretch the edges out in 1 pass, instead of doing a seperate pass for every level?
[/ QUOTE ]
The process have to repeat "radius" times. For example, for a radius of 8 it needs to repeat the process 8 times.
It also filters each map individually to get better results and to consume less memory. That's why you see the progress bar to restart the process.
I could detect the edges and dilate them only, yep. That's in fact a very clever idea. thx!
Just uploaded small patch solving some mouse problems + a small bug in the batch protection queue. Feel free to download it again in case you had problems with that, sorry!
i'll have the normals calculated in a matter of seconds and it will take twice as long just to do the 8 passes of dilation on a large map like a 2048. Also sometimes it will go through this process 2 or 3 times. Is there any way you go just stretch the edges out in 1 pass, instead of doing a seperate pass for every level?
[/ QUOTE ]
The process have to repeat "radius" times. For example, for a radius of 8 it needs to repeat the process 8 times.
It also filters each map individually to get better results and to consume less memory. That's why you see the progress bar to restart the process.
I could detect the edges and dilate them only, yep. That's in fact a very clever idea. thx!
[/ QUOTE ]
Yeah i understand that part, but sometimes with cetain models it will go through 1-8, and then have to go through 1-8 again, and even a third time. Its really a tribute to the app if the slowest thing it does is expand the pixels hehe.
First off amazing app. I have the same issue in the viewer with the most recent download. I can not see my models. If I turn on cage, normals etc they show up. The ui and the grid displays fine. I am running on a laptop with an ATI Mobility Radeon X1400 with all the latest
[ QUOTE ]
I have the same issue in the viewer with the most recent download. I can not see my models. If I turn on cage, normals etc they show up. The ui and the grid displays fine. I am running on a laptop with an ATI Mobility Radeon X1400 with all the latest
[/ QUOTE ]
You are using the DX9 graphics driver, aren't you? Under OpenGL should work ok. The new 3.11.0 will have that problem patched, don't worry ( the problem was that the SM3.0 instruction limit is just 512 on ATI hardware and the shaders were taking 660! )
That was it. Sorry I am an Xnormal noob. I just figured out how to change my drivers and it works great.
A couple of feature requests, as if you don't have enough going on already. Sorry if any of this already exists. I have gone through it all but I not run across these.
1. Edge padding. It would be great to have an edge padding distance the user can control.
2. Render to AVI. It would be really great to have the rotation of the model in the viewer render to an avi.
3. Mesh information - Poly count, etc
Either way wonderful app. Very inmpressive.
EDIT - Also it looks like the Normal to Cavity map hangs now on my systems. I have tried it on both computers at home and my one at work. It displays the progress window and says calculating cavity map but nothing happens. Even the abort button seems to not register
[ QUOTE ]
1. Edge padding. It would be great to have an edge padding distance the user can control
[/ QUOTE ]
You can! Just do a mouse right click over the "dilation" filter and "configure" it.
[ QUOTE ]
2. Render to AVI. It would be really great to have the rotation of the model in the viewer render to an avi.
[/ QUOTE ]
You can rotate the model/light automatically checking the "reelC" and "reelL" inside the viewer. About the AVI capture I was trying to implement it but unfortunally OpenGL/Direct3D takes too much time to pass the framebuffer to memory, rescale, compress it and store to disk... The best result I got was 5FPS capture and the video output so decided not to include it(too "jerky" animation result).
To get decent results I bet you will need to connect the video output to the video in of a videocapture card on other computer.
Also got other problem with propietary codecs(divX,quicktime,blah blah). Cannot include/use them without authorization( and they asked me for a lot $$$!). I can only use the ones with Windows comes with (Indeo,etc) which but that generates a too-big temporary file or Theora Ogg(pre-release alpha... not stable)
[ QUOTE ]
3. Mesh information - Poly count, etc
[/ QUOTE ]
You can see the # of vertices and faces inside the 3d viewer ( the model radius too, VRAM used, etc ). Yep, perhaps I should put it too in the mesh list.
[ QUOTE ]
Also it looks like the Normal to Cavity map hangs now on my systems. I have tried it on both computers at home and my one at work. It displays the progress window and says calculating cavity map but nothing happens. Even the abort button seems to not register
[/ QUOTE ]
The photoshop filter or the tool inside the xNormal.exe? Are you using the latest 3.10.6 version, don't you? Are you using Windows Vista x64?
Are you sure it hangs? Try to reduce the radius parameter...(put it to one to test). If you set it too high can take some time to update the progress bar.
If you have a dual core remember to install the WindowsXP dual-core patch(+SP2) and the AMD patch too ( in case you own an AMD X2 )
I just uploaded the 3.11.0 Beta 1.
Some bugs corrected ( including the hangings in the cavity map tool ) + some new features.
Btw, if you want to test the new hxGrid distributed cluster rendering system sure you have Windows XP x86( with the SP2 ). At the moment it doesn't work on Windows 2000/Vista ( we're still working to solve some problems ).
Basically you need to put the "coordinator" in one computer. Then you execute some "agents" in the different computers. Then you run xNormal and use the "agent computers" to render your map in one second You can install the "agents" in all the computers of your office... it only activates when the computer is idle so won't affect the speed of the machine in normal working mode.
For advanced configutations you need to edit the hxGrid.ini file ( but caution! don't modify it unless you are an advanced network administrator! )
If you have doubts about how to configure hxGrid please tell me and I will explain a bit more.
thx!
ps: I had to remove old versions due to excesive disk space occupied in the server, sorry!
[ QUOTE ]
By the way Jogshy, could you some day include alpha map display support into xNormal? I'd love to take all my pimping screens in it.
[/ QUOTE ]
Inside the 3D viewer? Just use the "transparency mode" in the 3D viewer tab after selecting a mesh... And put the transparency into the alpha channel of the base texture... See the Bojana's or MoP example for more info! Alpha-blended transparency sorting cannot be HW-accelerated efficiently ( so you need to specify a good mesh order drawing moving up/down meshes in the list ). Additive transparency does not need sort and neither 1bit one(colokey).
On the other hand, if you are referring to see the alpha channel of a bitmap or texture, just go to the "tools" tab and open the height map to normal map converter. Load an image and toggle the display alpha mode with the 3-rgb-lines / white small circle button.
Haha, I guess it should. I was just desperately looking for something like "alpha map" section in the 3D viewer menu without realizing it was under a different name. Will check it out, thanks a lot.
I'd never really paid any attention to this (as the normal map tools in Maya did me just fine) but I think the distributed rendering has definitely turned my eye. Since we've got a couple spare render machines, one a dual CPU, dual core machine Plus the ability to plug in your own tangent generation? Oh I can't wait to play! Thanks jogshy!
[ QUOTE ]
would it be possible to have the light position locked to the camera?
[/ QUOTE ]
Ok, will add a small button to do this in the final version.
[ QUOTE ]
I think the distributed rendering has definitely turned my eye. Since we've got a couple spare render machines, one a dual CPU, dual core machine Plus the ability to plug in your own tangent generation?
[/ QUOTE ]
Initially I am distributing the rendering process with hxGrid. At the moment works only with WinXP and x86(this means no W2000 or vista) but Roman is trying to port it to other OS in a few ( will wait for that before to release the final version I think ). Remeber to unblock the agent, coordinator and xNormal from the firewall or won't work well!
I am also considering other distributed renderer forms like MPI(lala BlueGene) or a custom service with IP sockets.
About the custom tangent, yep, you can use the xNormal SDK to import your own tangent basis or to export the calculated in xNormal one to a file or to use the xNormal SDK tangent basis computation routines directly in your 3dsmax/Maya/XSI/LW exporter. The guys at 8monkeylabs chose the first method ( to write a mesh importer for xNormal and use the tangent basis present on the Stooge mesh format ). Very flexible as you can see.
Hmm it seems the distributed render doesn't work for me.
I installed xNormal, with agent & distributed render plugin on my machine.
Installed it with co-ordinator & agent on the beefiest machine.
Installed it with agent on the 3rd machine.
I run the coordinator and see all three machines (and 45 gHz of computing power... sweet!) hooked up and ready to go.
On my machine, I load the flag example, hit render and I get: "Unexpected exception in the generation thread. Can't generate normal map". If I switch the renderer to the scanline, it generates just fine. Any ideas?
Installed it with co-ordinator & agent on the beefiest machine.
[/ QUOTE ]
Install the coordinator alone better. Will use a lot of resources. If you add an agent too to that machine will eat the RAM.
[ QUOTE ]
I installed xNormal, with agent & distributed render plugin on my machine.
[/ QUOTE ]
Don't run a agent in the xNormal machine. That machine needs tons of CPU available to receive and process the incomming data.
[ QUOTE ]
"Unexpected exception in the generation thread. Can't generate normal map". If I switch the renderer to the scanline, it generates just fine. Any ideas?
[/ QUOTE ]
Sure you install the coordinator and agent in a directory with All/All rights and with read/write permissions. If not, won't be able to read/write into its \TEMP folder. Don't need to share the folder, just sure the rights are ok.
Sure you unlock the xNormal.exe, and coordinator.exe, agent.exe on the Windows firewall. When the firewall dialog asks you click over the "Unlock", not over the "Ask me later" or "Deny".
Currently there is a bug in hxGrid that makes to crash in computers with more than 3Gb of RAM btw.
Finally, hxGrid only works at the moment on WindowsXP SP2 for x86. If you use WinNT 4.0, Win95/98/ME, Windows 2000, Windows Server 2003, Windows Vista or WinXP Pro 64 bits won't work. We are working to port it to other OS, is just question of time.
If is nothing all the mentioned works... well... it's a beta
Idea - Would it be possible to have a camera auto center on the models? So that when you have it do the rotation in the 3d view it auto rotates perfectly around the center of the model no matter how you rotate the camera or dolly it in and out? Just an idea.
Hmmm no luck. Reinstalled everything on network drives to be safe and we don't use Windows firewall here so no problems there. All machines are XP SP2 as well. No machine has more than 3 gigs of RAM either. I tried many combinations of which machines are the coordinator, if the agent is on the machine launching the task, or on the coordinator's machines, etc. Same old crash. Is there a debug log output anywhere to look through, I didn't see one at first glance?
[ QUOTE ]
Hmmm no luck. Reinstalled everything on network drives to be safe and we don't use Windows firewall here so no problems there. All machines are XP SP2 as well. No machine has more than 3 gigs of RAM either. I tried many combinations of which machines are the coordinator, if the agent is on the machine launching the task, or on the coordinator's machines, etc. Same old crash. Is there a debug log output anywhere to look through, I didn't see one at first glance?
[/ QUOTE ]
Do you see all the agents well in the coordinator window? Can you post a screenshot with the coordinator window,pls?
Crashes almost at end or when you start the map generation? Crashes too if you only mark the generate AO option?
Sure the MapRenderer_hxGrid_taskDLL.dll and the hxGridUserDLL.dll, zlib.dll are in the same folder than the xNormal.exe for x86. If not, sure you mark the hxGrid distributed render plugin in the installer(or use the zipped install)
Unfortunally there is no debug log for that at the moment, sorry.
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 ]
Idea - Would it be possible to have a camera auto center on the models?
Sorry if this has been answered. I looked through most of the thread and the only answer I can find is max does not export OBJ's well. It seems like along the seams of my UV's I am gettingsort of a disconnect. If I turn off the dilation I get a black line running around my UV seams. If I turn it on and up it almost seems to blur into my UV area, not just extend it out. This creates and almost blue, lack of normal detail, all along my UV seams. I must be missing something obvious. I am trying to Normal a head that us mirrored. I UV'd half of it. Mirrored it over and then moved and shrunk the mirrored UV's to their own little spot on the map. Is this a bad process?
EDIT - Well it is clearly an OBJ issue. I import my max exported OBj into Zbrush. Export that back out and bring it into Xnormal and I get much different results. Still some onomolies, but different results. Is there a way to "clean" obj's because I unwrap in max?
Thanks
Jesse
Yup, others have had that problem, and Jogshy made a new export format just for Max, called SBM. It's linked to earlier in the thread. Get it from his site, and only import SBMs instead of OBJs.
[ QUOTE ]
Sorry if this has been answered. I looked through most of the thread and the only answer I can find is max does not export OBJ's well. It seems like along the seams of my UV's I am gettingsort of a disconnect.
[/ QUOTE ]
The 3dsmax2obj mesh exporter that comes with 3dsmax is a bit old and buggy, yep. The 3DS can be also bad because it averages the vertex normals. The LWO does not save vertex normals.
To export well your smoothing groups and normals intact I recommend you to use the ASE or COLLADA mesh exporter(that comes with 3dsmax) or the xNormal SBM exporter as Poop said.
Also remember to resetXForm and collapse modifiers(this last is not always possible so if not, apply a edit mesh modifier in the top of the stack) before to export from 3dsmax.
thanks Jogshy for creating xNormal! the new cavity maps are impressive and quite useful.
one thing i also tried with xNormal is ambient occlusion and found the results a little wanting and could be better. i'm assuming its work in progress. i tried another AO tool called 'faogen' and the results are impressive. i'm hoping a future build of xNormal will rival the quality and speed of 'faogen' in terms of AO map generation.
yeah i have to say that AO is the only thing that is making me switch apps for the casting process. i really wish the AO maps were of higher quality. i always get tons of artifacts and errors with rendering them in Xnormal. i am currently having to use turtle at work to render my ambient occlusion maps and turtle is a pain to use.
[ QUOTE ]
What happened to the swizzle settings in v3.11.0.7820 beta 1? Am I missing something?
[/ QUOTE ]
Which ones? the normal map, the ao bent normals, the viewer3d or the tools one?
The normals one are in the new normal map dialog( press the "..." near the "generate normal and displacement map" ). The AO-bent normals one in the "..." of the "generate ambient occlusion map".
The viewer ones just appear when you select a normal map file in the slot.
[ QUOTE ]
i really wish the AO maps were of higher quality. i always get tons of artifacts and errors with rendering them
[/ QUOTE ]
Remember to control the AO bias, aperture angle and use the cages to limit the distance of the rays( or, on the contrary, set the "don't limit ray distance" option and see if that solves it). If you can post any photo and settings I can tell you more!
Replies
Jogshy did you get the code for the stooge .mesh importer btw?
[/ QUOTE ]
Shhhhh! don't ruin the surprises (yes, I did :P)
http://www.xnormal.net
I changed a lot the CPU multicore code. Please tell me if you find any error/bug/crash!
The final version will include a lot of new things, but I need to be sure first that the new multicore renderer works ok! So that's why a beta was released.
thx, all feedback will be welcome!
- New render AO without highpoly feature with normal map support ( specially dedicated to Johny! ) Select your lowpoly model in both the highpoly/lowpoly grid, specify a normal map(optional) and you will get fast and cheap AO approximation ( and remember to use cages to limit distance for better results ). Also can be used to get more detail in the default AO system. What does is to invalidate the hit surface normal and use the normal present in the normal map.
- Two new tools: height map to occlusion map + PhotoNormal to make normal maps directly from photographs!
- Copy/paste support in the tools
- New normal-to-cavity methods
- Added antialiased/non-antialiased map mask output files
- Added two new tangent basis calculators(D3DX,MeshMender) to improve compatibility with programs and 3d engines.
- Added the 8MonkeyLabs mesh importer for their upcoming game Darkest of days ( I heard this game gonna revolutionise the FPS shooter genere! )
- Re-structured the program internally to be more memory efficient and also coded again the multicore/multithread routines( I fear that, pls let me know if you have any crash problem related to this ).
- Added camera/light automatic movement inside the 3D viewer
- Solved multiple bugs ( Photoshop filters displaced bitmap preview, grayscale JPG import error, OpenGL window resize, interior AO, incorrect external cage warning, etc... ) and also i'm pretty sure I added some new bugs too
Btw, requires the latest June 2007 DirectX, sorry I'm a M$ofer
ps: Already working in the 3.11.0 which gonna include a new and amazing fast renderer. But that will be after the summer, vacations are just here !
See ya!
going to try the AO from lowpoly with normals now
its amazing your commitment for this life saviour
the only issue i have is with the cavity mapping tool, became too confusing and takes along time to generate anything
[/ QUOTE ]
The EPD method is the fastest. For the Min,max,avg,spd try to reduce the search radius from 8 pixels to 3/4. (Min/Max are strange, use avg/spd usually ).
If you have a lowpoly mesh + a normal map better use the new render AO with normal map feature, probably will give you better results.
About the render speed I think i can improve it a lot ( specially the AO). In fact, the new 3.11 i'm preparing will smash all you've seen!
hey jogshy, another thing, the 3d viewer doesnt work for me, i mean it works, but i cant use the mouse, the cursor gets stuck in the upper corner
[/ QUOTE ]
Are you using the windowed mode or full screen? If you are using windowed mode remember to ALT+ENTER to release/capture the mouse cursor.
Also try to center the window, there is a bug currently with the mouse cursor if the window is clipped( where 'clipped' means you move a part the window outside the monitor). That will be patched shortly.
One thing to consider is that pixels of previous/next row are "far away" in the memory, which makes things slower.
btw awesome work jogshy, that you continue your perfect support for this tool for so long, is really top notch. I hope some of you pros using this tool frequently remember to "donate" something back... Yes it's for free, but if you use it to make money, you should not forget the immense workload people put in free software...
i'll have the normals calculated in a matter of seconds and it will take twice as long just to do the 8 passes of dilation on a large map like a 2048. Also sometimes it will go through this process 2 or 3 times. Is there any way you go just stretch the edges out in 1 pass, instead of doing a seperate pass for every level?
[/ QUOTE ]
The process have to repeat "radius" times. For example, for a radius of 8 it needs to repeat the process 8 times.
It also filters each map individually to get better results and to consume less memory. That's why you see the progress bar to restart the process.
I could detect the edges and dilate them only, yep. That's in fact a very clever idea. thx!
[ QUOTE ]
i'll have the normals calculated in a matter of seconds and it will take twice as long just to do the 8 passes of dilation on a large map like a 2048. Also sometimes it will go through this process 2 or 3 times. Is there any way you go just stretch the edges out in 1 pass, instead of doing a seperate pass for every level?
[/ QUOTE ]
The process have to repeat "radius" times. For example, for a radius of 8 it needs to repeat the process 8 times.
It also filters each map individually to get better results and to consume less memory. That's why you see the progress bar to restart the process.
I could detect the edges and dilate them only, yep. That's in fact a very clever idea. thx!
[/ QUOTE ]
Yeah i understand that part, but sometimes with cetain models it will go through 1-8, and then have to go through 1-8 again, and even a third time. Its really a tribute to the app if the slowest thing it does is expand the pixels hehe.
I have the same issue in the viewer with the most recent download. I can not see my models. If I turn on cage, normals etc they show up. The ui and the grid displays fine. I am running on a laptop with an ATI Mobility Radeon X1400 with all the latest
[/ QUOTE ]
You are using the DX9 graphics driver, aren't you? Under OpenGL should work ok. The new 3.11.0 will have that problem patched, don't worry ( the problem was that the SM3.0 instruction limit is just 512 on ATI hardware and the shaders were taking 660! )
A couple of feature requests, as if you don't have enough going on already. Sorry if any of this already exists. I have gone through it all but I not run across these.
1. Edge padding. It would be great to have an edge padding distance the user can control.
2. Render to AVI. It would be really great to have the rotation of the model in the viewer render to an avi.
3. Mesh information - Poly count, etc
Either way wonderful app. Very inmpressive.
EDIT - Also it looks like the Normal to Cavity map hangs now on my systems. I have tried it on both computers at home and my one at work. It displays the progress window and says calculating cavity map but nothing happens. Even the abort button seems to not register
1. Edge padding. It would be great to have an edge padding distance the user can control
[/ QUOTE ]
You can! Just do a mouse right click over the "dilation" filter and "configure" it.
[ QUOTE ]
2. Render to AVI. It would be really great to have the rotation of the model in the viewer render to an avi.
[/ QUOTE ]
You can rotate the model/light automatically checking the "reelC" and "reelL" inside the viewer. About the AVI capture I was trying to implement it but unfortunally OpenGL/Direct3D takes too much time to pass the framebuffer to memory, rescale, compress it and store to disk... The best result I got was 5FPS capture and the video output so decided not to include it(too "jerky" animation result).
To get decent results I bet you will need to connect the video output to the video in of a videocapture card on other computer.
Also got other problem with propietary codecs(divX,quicktime,blah blah). Cannot include/use them without authorization( and they asked me for a lot $$$!). I can only use the ones with Windows comes with (Indeo,etc) which but that generates a too-big temporary file or Theora Ogg(pre-release alpha... not stable)
[ QUOTE ]
3. Mesh information - Poly count, etc
[/ QUOTE ]
You can see the # of vertices and faces inside the 3d viewer ( the model radius too, VRAM used, etc ). Yep, perhaps I should put it too in the mesh list.
[ QUOTE ]
Also it looks like the Normal to Cavity map hangs now on my systems. I have tried it on both computers at home and my one at work. It displays the progress window and says calculating cavity map but nothing happens. Even the abort button seems to not register
[/ QUOTE ]
The photoshop filter or the tool inside the xNormal.exe? Are you using the latest 3.10.6 version, don't you? Are you using Windows Vista x64?
Are you sure it hangs? Try to reduce the radius parameter...(put it to one to test). If you set it too high can take some time to update the progress bar.
If you have a dual core remember to install the WindowsXP dual-core patch(+SP2) and the AMD patch too ( in case you own an AMD X2 )
Well, gonna revise it a few just in case!
Got the same problem... older Cavity map versions work, but the 3.10.6 version hangs every time.
[/ QUOTE ]
The others work ok? I mean the heightmap to normal map, etc. Could you tell me too if the cavity map for Photoshop works ok too pls?
what CPU and OS do you use? The problem is related with multicore I bet cuz I changed all the multicore programming in the 3.10.6.
Well, i'm gonna post the 3.11.0 beta 1 in a few with new features and some bugs corrected.
Some bugs corrected ( including the hangings in the cavity map tool ) + some new features.
Btw, if you want to test the new hxGrid distributed cluster rendering system sure you have Windows XP x86( with the SP2 ). At the moment it doesn't work on Windows 2000/Vista ( we're still working to solve some problems ).
Basically you need to put the "coordinator" in one computer. Then you execute some "agents" in the different computers. Then you run xNormal and use the "agent computers" to render your map in one second You can install the "agents" in all the computers of your office... it only activates when the computer is idle so won't affect the speed of the machine in normal working mode.
For advanced configutations you need to edit the hxGrid.ini file ( but caution! don't modify it unless you are an advanced network administrator! )
If you have doubts about how to configure hxGrid please tell me and I will explain a bit more.
thx!
ps: I had to remove old versions due to excesive disk space occupied in the server, sorry!
By the way Jogshy, could you some day include alpha map display support into xNormal? I'd love to take all my pimping screens in it.
[/ QUOTE ]
Inside the 3D viewer? Just use the "transparency mode" in the 3D viewer tab after selecting a mesh... And put the transparency into the alpha channel of the base texture... See the Bojana's or MoP example for more info! Alpha-blended transparency sorting cannot be HW-accelerated efficiently ( so you need to specify a good mesh order drawing moving up/down meshes in the list ). Additive transparency does not need sort and neither 1bit one(colokey).
On the other hand, if you are referring to see the alpha channel of a bitmap or texture, just go to the "tools" tab and open the height map to normal map converter. Load an image and toggle the display alpha mode with the 3-rgb-lines / white small circle button.
Hope it helps.
Hope it helps.
[/ QUOTE ]
Haha, I guess it should. I was just desperately looking for something like "alpha map" section in the 3D viewer menu without realizing it was under a different name. Will check it out, thanks a lot.
When I alt-tab out of the 3d Viewer and comeback the screen hangs. Otherwise another bang-up job thanks,
When I alt-tab out of the 3d Viewer and comeback the screen hangs.
[/ QUOTE ]
Ouch, true... Is solved now and will include that for the final version.
Anybody detected any other bug or have any other suggestion/comment before I release the final version, please?
In the 3d viewer, when the camera and light rotate, would it be possible to have the light position locked to the camera?
would it be possible to have the light position locked to the camera?
[/ QUOTE ]
Ok, will add a small button to do this in the final version.
[ QUOTE ]
I think the distributed rendering has definitely turned my eye. Since we've got a couple spare render machines, one a dual CPU, dual core machine Plus the ability to plug in your own tangent generation?
[/ QUOTE ]
Initially I am distributing the rendering process with hxGrid. At the moment works only with WinXP and x86(this means no W2000 or vista) but Roman is trying to port it to other OS in a few ( will wait for that before to release the final version I think ). Remeber to unblock the agent, coordinator and xNormal from the firewall or won't work well!
I am also considering other distributed renderer forms like MPI(lala BlueGene) or a custom service with IP sockets.
About the custom tangent, yep, you can use the xNormal SDK to import your own tangent basis or to export the calculated in xNormal one to a file or to use the xNormal SDK tangent basis computation routines directly in your 3dsmax/Maya/XSI/LW exporter. The guys at 8monkeylabs chose the first method ( to write a mesh importer for xNormal and use the tangent basis present on the Stooge mesh format ). Very flexible as you can see.
I installed xNormal, with agent & distributed render plugin on my machine.
Installed it with co-ordinator & agent on the beefiest machine.
Installed it with agent on the 3rd machine.
I run the coordinator and see all three machines (and 45 gHz of computing power... sweet!) hooked up and ready to go.
On my machine, I load the flag example, hit render and I get: "Unexpected exception in the generation thread. Can't generate normal map". If I switch the renderer to the scanline, it generates just fine. Any ideas?
Installed it with co-ordinator & agent on the beefiest machine.
[/ QUOTE ]
Install the coordinator alone better. Will use a lot of resources. If you add an agent too to that machine will eat the RAM.
[ QUOTE ]
I installed xNormal, with agent & distributed render plugin on my machine.
[/ QUOTE ]
Don't run a agent in the xNormal machine. That machine needs tons of CPU available to receive and process the incomming data.
[ QUOTE ]
"Unexpected exception in the generation thread. Can't generate normal map". If I switch the renderer to the scanline, it generates just fine. Any ideas?
[/ QUOTE ]
Sure you install the coordinator and agent in a directory with All/All rights and with read/write permissions. If not, won't be able to read/write into its \TEMP folder. Don't need to share the folder, just sure the rights are ok.
Sure you unlock the xNormal.exe, and coordinator.exe, agent.exe on the Windows firewall. When the firewall dialog asks you click over the "Unlock", not over the "Ask me later" or "Deny".
Currently there is a bug in hxGrid that makes to crash in computers with more than 3Gb of RAM btw.
Finally, hxGrid only works at the moment on WindowsXP SP2 for x86. If you use WinNT 4.0, Win95/98/ME, Windows 2000, Windows Server 2003, Windows Vista or WinXP Pro 64 bits won't work. We are working to port it to other OS, is just question of time.
If is nothing all the mentioned works... well... it's a beta
Idea - Would it be possible to have a camera auto center on the models? So that when you have it do the rotation in the 3d view it auto rotates perfectly around the center of the model no matter how you rotate the camera or dolly it in and out? Just an idea.
Hmmm no luck. Reinstalled everything on network drives to be safe and we don't use Windows firewall here so no problems there. All machines are XP SP2 as well. No machine has more than 3 gigs of RAM either. I tried many combinations of which machines are the coordinator, if the agent is on the machine launching the task, or on the coordinator's machines, etc. Same old crash. Is there a debug log output anywhere to look through, I didn't see one at first glance?
[/ QUOTE ]
Do you see all the agents well in the coordinator window? Can you post a screenshot with the coordinator window,pls?
Crashes almost at end or when you start the map generation? Crashes too if you only mark the generate AO option?
Sure the MapRenderer_hxGrid_taskDLL.dll and the hxGridUserDLL.dll, zlib.dll are in the same folder than the xNormal.exe for x86. If not, sure you mark the hxGrid distributed render plugin in the installer(or use the zipped install)
Unfortunally there is no debug log for that at the moment, sorry.
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 ]
Idea - Would it be possible to have a camera auto center on the models?
[/ QUOTE ]
Ok
EDIT - Well it is clearly an OBJ issue. I import my max exported OBj into Zbrush. Export that back out and bring it into Xnormal and I get much different results. Still some onomolies, but different results. Is there a way to "clean" obj's because I unwrap in max?
Thanks
Jesse
Sorry if this has been answered. I looked through most of the thread and the only answer I can find is max does not export OBJ's well. It seems like along the seams of my UV's I am gettingsort of a disconnect.
[/ QUOTE ]
The 3dsmax2obj mesh exporter that comes with 3dsmax is a bit old and buggy, yep. The 3DS can be also bad because it averages the vertex normals. The LWO does not save vertex normals.
To export well your smoothing groups and normals intact I recommend you to use the ASE or COLLADA mesh exporter(that comes with 3dsmax) or the xNormal SBM exporter as Poop said.
Also remember to resetXForm and collapse modifiers(this last is not always possible so if not, apply a edit mesh modifier in the top of the stack) before to export from 3dsmax.
Hope if helps.
one thing i also tried with xNormal is ambient occlusion and found the results a little wanting and could be better. i'm assuming its work in progress. i tried another AO tool called 'faogen' and the results are impressive. i'm hoping a future build of xNormal will rival the quality and speed of 'faogen' in terms of AO map generation.
cheers,
Rom
What happened to the swizzle settings in v3.11.0.7820 beta 1? Am I missing something?
[/ QUOTE ]
Which ones? the normal map, the ao bent normals, the viewer3d or the tools one?
The normals one are in the new normal map dialog( press the "..." near the "generate normal and displacement map" ). The AO-bent normals one in the "..." of the "generate ambient occlusion map".
The viewer ones just appear when you select a normal map file in the slot.
[ QUOTE ]
i really wish the AO maps were of higher quality. i always get tons of artifacts and errors with rendering them
[/ QUOTE ]
Remember to control the AO bias, aperture angle and use the cages to limit the distance of the rays( or, on the contrary, set the "don't limit ray distance" option and see if that solves it). If you can post any photo and settings I can tell you more!