I’m thrilled at the prospect of implementing an ACES workflow for an upcoming CG short. I am having some trouble with the actual execution, however.
Using Maya, Renderman 23, and finishing in Resolve / Fusion.
You can see below that I can’t get a texture to render correctly to match in ACEScg.
I have a source image made in photoshop in sRGB, which I’ve then converted using Utility - sRGB - Tex. That’s the middle image, but… it’s rendering too dark in comparison.
(Also including the raw sRGB texture, linearized, and as expected it’s too saturated.)
Using no texture but only Maya’s color picker with the same RGB values as photoshop, and that is a very close match.
Why is my converted ACEScg texture rendering darker than expected?
I’m saving the texture as an sRGB TIFF file from photoshop originally. Converting using this online conversion website.
Even after reading through some of the posts here, I’m struggling in general with photoshop and ACES (seems really painful to set up and work with photoshop in aces), and dreading approaching Substance.
Is there also a way to batch convert textures to ACES? I don’t have a Nuke license.
Hopefully Renderman will implement some of this in the upcoming 24, since it doesn’t include a way to convert textures in their ‘file’ node, and color temp / skylight are still not set up correctly…
I have a small tool for batch conversion: https://gum.co/pycocs
Then I think we miss some precisions about your worflow, what does the colorpicker in Maya looks like (what settings), what are you picking,… ? There is a lot of variables.
One of the option is simply to eyeball the color you want to reproduce while working under the ACES ODT (ie in the viewport)
Textures look darker because you are displaying them through the ACES Output Transform in Maya.
I suspect you are using the default ACES OCIO Config where the color picking role is set to Output_sRGB. Hence the close match when you are using the Color UI.
Chris, you’re absolutely correct on both counts and after more carefully reading that section of your (amazing, really helpful guide) I do understand these behaviors. And Liam- thank you, your tool looks great!
I do have two questions about this that I’m still unsure of from the book.
Just to really understand, why exactly do converted sRGB textures always get darker? If it’s a change because of the RRT / ODT ‘filmic s curve’, shouldn’t some brighter values actually be mapped to a brighter value? EG, if something is already towards the mids shoulder of the curve and I add an s curve, shouldnt it become slightly brighter? This isn’t super important, just curious - sure I just don’t fully understand everything. I see why this works physically (the base color wouldn’t be seen as that exact pure base color in a scene referred image, makes sense).
After understanding this color shift, I’m still having trouble with the technical execution of matching a painted sRGB concept image with sRGB textures. In your book (you have an example with the steamboat) you bring the sRGB image into nuke using the OUT conversion method, then sample the colors to get the corresponding correct value. Then you use those to shade / color the model.
If i’m in Substance Painter and using those sampled values to paint the texture… isn’t this the same thing everyone says we shouldn’t do- basically using colors from the OUT sRGB as an input? We’ve just manually gone in and done the conversion by painting, then reading it with Utility sRGB Tex.
Just curious how this actually gets done in practice on fully CG shows with painted textures.
Thanks for the help though guys, I do now understand what’s happening with my original posted image!
No, a s-curve should not make things “brighter”. It is used to map luminance values onto display (from scene-referred to display-referred). It generally adds contrast with a nice roll-off in the highlights. This is very well explained in Cinematic Color.
You’re absolutely right. It is not recommended at all to load texures in Output - sRGB. It has been discussed many times on ACESCentral. However in my boat example, I am loading a color key to approximately pick some ACEScg values and then use them in Guerilla. Since I am hand-picking them, I can check they are within a reasonable PBR range.
I think this example was to show a workflow that would hopefully still be PBR-friendly and Art Director friendly. Maybe it is a flawed process and an unsuccessful experiment, but it kinda worked for me and my Playmobil boat.
