How do I correctly import srgb tetxures into houdini arnold and nuke?

Hello everybody,

I’m working on a 3d stylized animation project. I got srgb textures as pngs and some of them have 1/1/1 whites. When I use the import transform utility-srgb-texture in arnold for houdini or even nuke the whites are 1/1/1, but look grey viewing them through the ACES 1.0 -SDR Video (sRGB-Display) ODT. Using Un-tone-mapped (sRGB-Display) they look identical to the pngs. Should I now apply an inverted ODT on these textures in nuke by using the OCIODisplay node to get back these whites giving me 16/16/16 or should I increase the lighting in the scene or make the textures brighter in houdini by just eyeballing. Or is it correct I never see pure white 16/16/16 purely from textures? Only by increasing the light exposure? Or should I treat my textures like logos that have to show up correctly even when viewed through the ODT? Whats the correct way?

Thanks in advance for your help!
Cheers!
Chanse

Hello Chanse,

it depends a bit on your project but I would recommend this :

About the second question:

yes, this is correct. Textures should be within [0,1] range.

I would not recommend this for PBR:

Regards,
Chris

Thanks for your answer!! The project mostly uses NPR. What would you suggest, if I have to closely hit the colors of concept art?

The one-million dollar question :wink:

I think you have two paths ahead of you :

  • Either you need to closely hit the concept art (almost like pixel matching) and then maybe you can use the display encoding view (un-tone-mapped). Almost like a 2d project with no Global Illumination (GI) where everything is painted (almost with emissive materials).
  • Either you have bit of room at shot lighting/rendering stage with the GI and some Comp work. I have heard of NPR projects using color management but there was definitely room for shot improvements/proposals, rather than matching perfectly a 2d reference.

This is why my first answer was “it depends on the project” like time/budget… Thanks !

Bear in mind that while it’s a valid creative choice, replicating textures with (1, 1, 1) whites in the final rendered image means you have nowhere brighter to go for specular highlights. That’s the reason systems like ACES render diffuse white at a lower level, to allow room above that for highlight roll-off.

Adding to @nick 's point here. Likewise if you take a picture of a painting on a white canvas, the white of the canvas is not pure white. If it was you could not see the canvas fibers. It’s the classic “polar bear in a snow storm” dilemma.

A word of caution with rendering with emissive materials. In a physically based material emissive materials will affect the lighting and reflection of objects in unwanted ways for an NPR project. I would suggest instead a combination approach of rendering the “ambient light” with a dome light in the 3D rendering app – the purpose here being that artists can have a reasonable aproximation of the final look in the 3D app, and then in compositing replacing this with a modified albedo AOV for the final look of film. In general, in NPR productions, compositing needs to play a major role, as this is where you need to rebuild the AOVs so they do not assembled in a physically realistic fashion, but rather in one that reflects the stylized, artistic aims of the show. Here, you are essentially intentionally breaking all the rules of the physically based render to achieve the NPR result.