Render passes for HDRI?

[magneto]

Hi,

 

Anyone knows what passes should I render for HDRI renders? The 3dbuzz HDRI tutorial shows setting up how to use HDRI in a shader and renders a separate image using the shader with the Occlusion VOP without an environment map (1), another Occlusion VOP with the diffuse environment map (2), and another one using the Environment Map VOP with the same diffuse environment map (3).

 

Then there is the pure color pass with no lighting or shadows (4).

 

The workflow shown in the tutorial seems suboptimal/outdated though (Houdini 9). He doesn't combine everything 1:1 but blends (1) with (2) partially and then adds (3) partially to the result.

 

Is this the best way to do this? I looked online for the right passes to render with other 3d software but couldn't find anything useful that mentions the right passes to bring together.

 

 

Thanks

[anim]

HDRI is just light so you can treat it as such

 

so if you are splitting your renders to diffuse, reflection, refraction, emission, you can as well split each of the components per light so that you can get HDRI diffuse, HDRI reflection, HDRI refraction

 

if you need more control you can split each of them to direct/indirect

 

at least that's quite common approach for physical renders (and they usually combine by pure addition in linear color space)

 

you can always output raw color if you really wish to tweak it a lot in comp

 

the point is you choose the passes you need there is no one "right passes" solution, since there are different degrees of freedom you want to have in comp, but you still want to avoid overcomplicating your setups

 

I find splitting renders into components or per light components sufficient for most of the comp work, with additional passes as you need (like depth, P/N, motion vectors, ...)

 

reflection/ambient occlusion are old school passes not necessary in physically based pipelines, however if you understand their main purpose, you can still take advantage of them mainly if you need fast renders and not necessarily aim for accurate physical lighting and reflections

Edited March 3, 2014 by anim

[magneto]

Thanks anim, that clarified things for me. Now I have different questions though

 

1. Is there a shader in Houdini that's designed to accept HDR images and output HDR related passes?

 

2. The author builds each component himself like diffuse, AO, AO with environment map, reflection, reflection occlusion, etc which are straightforward but I am not sure how to split them into direct and indirect inside the shader?

 

3. By raw color, do you mean the diffuse color pass without any lighting?

 

4. Is it reasonable to use an HDR image in your environment light instead when you don't need to do any compositing?

 

5. Is there a guide you know of that explains the most up to date rendering workflows in Houdini? I had no idea AO was not necessary nowadays, or that people now prefer PBR? I thought raytrace was the preferred method. I remember reading about Arnold renderer raytracing alot of things but I might be wrong. I am just not aware of the most established rendering workflows as the rendering tutorials for Houdini are quite old.

 

 

Thanks again

[anim]

1. any Surfce Model VOP / Mantra Surface shader based material + Environment Light with your HDR

 

2. use 1. and in Mantra output Combined Diffuse, Combined Reflection, ... AOVs for simpler passes

or

Direct Diffuse, Indirect Diffuse, Direct Reflection, Indirect Reflection, ... for more control in comp

 

if you have more lights, you can use Per Light exports for any (or all) of the AOVs so you have your HDRI light separate if you want

 

3. yes, just custom export variable for pure color without lighting

 

4. it's prefered to do that as described in 1. and2.  (not sure what else are you using, just HDRI map sampling in shader?)

 

5. not sure if such guide exists, it's mostly about following the latest development, but it's the mater of preference as well

reading houdini docs can definitely help

 

I didn't mention PBR, but in more general way physically based renderers, which includes mantra PBR, Arnold, Renderman's physically plausible shading, ...

they all use raytracing or pathtracing but in physically based way using energy conservation, BRDF shaders, ...

Edited March 3, 2014 by anim

[magneto]

Thanks alot anim, I understand it better now.

 

1. So basically no HDR in the shader but the light only? This might be outdated but the author was saying it's better to do all HDR sampling inside the shader because it's faster and you get away from samples for the diffuse. Also said you lose the ability to output different passes as the render becomes a single combined image. 

 

4. Yes basically using Environment Map VOP and Occlusion VOP with the HDR image and Occlusion VOP without an HDR image and just rendering to different passes.

 

Although the author had a really cumbersome workflow IMO where there was only one pass output each time you render and it would switch between different passes inside the shader using a Null ROP outside that the shaders are hard coded to look up for the current pass to render. So that it's all a single paratemer that controls all instances of this shader.

 

My understanding is that it's better to specify which passes you want in the renderer and output/export all passes inside the shader. Is this bad? Would Mantra render passes that you didn't specify in the Mantra ROP? I thought it would render the image and only execute the code paths in the shader that exports to your specified passes in the Mantra ROP.

 

That's what I want to do.

 

 

Thanks again