Texture Baking tutorials

[dennis]

thanks for you answer n6

the lightmap you posted here is generated by deep uv ?..hmm unfortunatly i cant get the demo to work on my machine.

maybe Vlado has a working version of bake3D for max 7? Vlado ?

well i dont have a problem with the overall look of the lightmap when using vray.
Using QMC as the first bounce it looks ok, although i have only tried with a test scene with one room if i have a more complex scene it sounds like a good idea to generate more IRmaps, thanks.

The biggest problem is that vray doesnt pad the edges of the lightmap probably it just draws a white edge if padding is above 0.

maybe its something i do wrong altough i cant really figure what it should be. What version do you use ?


a crop of the the rawGI output where you can see the white edge ( it should just have padded the color at the edge 1 pixel outwards )

[vlado]

The edge padding is correct, the problem is in the secondary rays bias. This has been corrected for later builds, but for now, you can just make the walls of your room thick, instead of a single plane.

Best regards,
Vlado

[dennis]

Thanks for your answer on this Vlado

The edge padding is correct..

can you elaborate a bit on this ? is it correct to pad with just a random color value ( white in this case ). should you not take the colorvalue in the lightmap at the border and pad that very same colorvalue x pixels outward, so edge padding is a post process.

..This has been corrected for later builds, but for now, you can just make the walls of your room thick, instead of a single plane.

Best regards,
Vlado

is 'light leaks' the reason ? thickness is not really a option as its realtime content ( or test to get an idea of workflow for ).
sounds good its corredted though so i will just patiently await the new version.

Thank you

[vlado]

You can put some shell around the walls while baking the light map. The shell doesn't have to go into the final geometry, and it doesn't need baking either.

As for the correctness, the edge padding represent the colors exactly on the border of the geometry (it is not done by duplication of pixels as this would interfere with antialiasing). The illumination at that point is not well defined. The image below is for normal lighting, and not for GI, but the principle is the same:



Point A is inside the floor surface and the renderer can correctly determine that it is in shadow. Point B however, is on the exact border of both surfaces. The shadow ray for that point doesn't hit anything, so the renderer considers it to be fully lit. This is perfectly correct, although it almost certainly is not what you want. The result is light leaks. This is not limited to texture baking, it can happen in normal rendering as well. To get rid of this ambiguity, you should make the walls with some thickness.

Like I said, newer versions handle this situation better, but there is no general method to avoid this effect other than to make the walls thick. This is also not limited to V-Ray, but to any renderer in general.

Best regards,
Vlado

[dennis]

i really appriciate you take the time to answer these simple issues and i am sorry to keep going but i just want you to understand me.

light leaks i understand is a difficult issue to deal with, but i dont think this is the issue here or at least is shouldnt be.

image below is with edge padding at zero and its actually very good no light leaks or anything even though the qmc settings is very low. if only i was able to pad the border by one pixel as a _post process_ not during light calc everything would be perfect


here is the same with edge padding at 2


btw. i tried the shell modifier and same thing.

Did you understand me the first time ?

[n6]

Can you send me your scene? I could give it a try if you want

[dennis]

that would be nice, have a look here

[vlado]

Did you understand me the first time ?

Yep, and I wanted to explain why you get this issue.

Best regards,
Vlado

[n6]

Vlado is right : its light leaks.
I will upload the result with "correct" geometry topology.

[dennis]

ok ok ok...i give up, im just that stupid.

i just cant understand why i have no _visible_ light leaks in the one without padding, and just to keep being stuborn ( wich is rather annoying combined with stupidity ) i would say that all the lightmap generators i've seen does padding as a post process, including 3ds scanline

looking forward to see your solution n6

and thanks for your patience both of you..

[n6]

Its still a buggy solution with a lot of leaks but at least the padding is now ok.
I have change other things in the rendering settings but it has nothing to do with the problem.

You "just" need to build an outside geometry to avoid those light leaks.









The scene :

http://www.visites-3d.com/visites-3d...m/lm_test2.max

[dennis]

thanks a lot n6 this reminds me of how i dealt with it with 3ds scanline and radiosity ( a draaaag )

it clearly shows its light leaks wich is evident when padding is on. my qustion is then how does the padding get calculated. is it done during the GI calculations by extending the polygon, somehow it looks like that to me as there where are no visible leaks when padding is off ( repeating myself i know )

nevermind though because we have a new improved version in a couple of days, right Vlado ?

[TRGraphics]

N6- What are your thoughts on Demicron WireFusion? I noticed that you have posted in the forum at their site with some very nice work.

I downloaded the demo and it seems very impressive. I played around with baking textures with vray but I can't seem to use vray materials to import into the program. Also, there seems to be a lot of bugs with vray and MAX's rtt exporter, crashes all the time.

Tony

[n6]

N6- What are your thoughts on Demicron WireFusion? I noticed that you have posted in the forum at their site with some very nice work.

Yep thanks Well first this not hardware accelerated its only a java based solution. But it is very intuitive and easy to use I think.

I downloaded the demo and it seems very impressive. I played around with baking textures with vray but I can't seem to use vray materials to import into the program. Also, there seems to be a lot of bugs with vray and MAX's rtt exporter, crashes all the time.

What are you calling "the rtt exporter"? Of course VRay material can't be exported by any web3d exporter, you need to convert all your materials (baked or not ) to standards materials.
Also 90% of the web3d plugins didnt support more than one UV set so if your baked material is, let say, in the UV channel 3, then it will not be exported.

Some exporter/viewers supports more than 1 UV set to do multitexturing for exemple diffuse texture with UV channel 1 , rawGI in channel 2, rawLighting in channel 3 etc... The BSExporter from http://www.bitmanagement.de do this.
But standard VRML exporter will not.

[TRGraphics]

Sorry, I meant to say the render to texture process not exporter. It crashes ever other time I try to bake the texture.

I'm really more concerned about quality of output than anything else. Do you feel it's as good as the others out there right now if you bake your textures? Baking textures seems to be the best solution for high quality output.