4.9 color grading change causing problems.

I recently upgraded to 4.9 and am sorting out the trailer load of rendering issues. The color grading or contrast has changed… somehow. All of my levels look wrong, shadowing has gone dark, colors and atmosphere are off and I can’t for the life of me get it back to how is is supposed to be.

I found a small line in the notes about some “corrections” made to color grading, which seems strange since everyone would have it tuned to how it was. I really need to know how to either undo these corrections if they are the cause, or offset them.

Thanks!

Hi docjor -

Just so I get a baseline, what version of the engine are you coming from 4.8 or 4.7?

Thank You

Eric Ketchum

Hey Eric, I’m coming from 4.8.

Hi docjor -

The line about Color Grading that is mentioned in the 4.9 release notes refers to a bug found in reflective surfaces when using Color Grading in 4.8. You could possible be seeing the switch from the Burley Diffuse Shading Model which was introduced in 4.8 and the Lambert Shading Model in 4.9, which was also in 4.7. You can read about it a little more on this post here:

Eric Ketchum

Aha, this may very well be the cause, though I would have to test it to be sure. Is there any way to change or test this in 4.9? And is there any changes like this one planned in 4.1 etc?

Thanks for the info, either way!

You can test it by going into ShadingModels.usf line 62 and adjusting the following:

	float3 Diffuse = Diffuse_Lambert( GBuffer.DiffuseColor );
	//float3 Diffuse = Diffuse_Burley( GBuffer.DiffuseColor, LobeRoughness[1], NoV, NoL, VoH );
	//float3 Diffuse = Diffuse_OrenNayar( GBuffer.DiffuseColor, LobeRoughness[1], NoV, NoL, VoH );

to

	//float3 Diffuse = Diffuse_Lambert( GBuffer.DiffuseColor );
	float3 Diffuse = Diffuse_Burley( GBuffer.DiffuseColor, LobeRoughness[1], NoV, NoL, VoH );
	//float3 Diffuse = Diffuse_OrenNayar( GBuffer.DiffuseColor, LobeRoughness[1], NoV, NoL, VoH );

Thanks! I made this change and rats it didn’t work. I’ll post a comparison SS to give you an idea of what I am seeing

Well, hmmm, can you post a picture of your Post Process Settings? I want to run some tests using your settings here.

Thank You

Eric Ketchum

Sorry for the so reply. I think this is all of it–

Hi docjor -

Thank you for your settings, that helped me track down the issue. I was right in my first guess that it is an issue of using Burley versus Lambert Diffuse Models. I am going to check with our engineers to see what else needs to be changed in code to successful switch back, as soon as I can confirm the switch I will right back with detailed instructions.

Thank you again -

Eric Ketchum

Hi docjor -

Thank you for your patience on this issue. After further investigation I was able to identify that while the change from Lambert to Burley might make a small difference, the biggest change was a correction to an issue involving color grading and skylights. In 4.8, Skylights were incorrectly distributing color grading and color grading itself had some issues because of the newly introduced tonemapper.

This issues were corrected in 4.9, but in your case you developed your map around those limitations and came up with a look that you enjoy. Without going back to 4.8, you will probably never be able to get the exact look that you setup in your project. That being said, you can play around with the r.Color.Max, r.Color.Mid and r.Color.Min settings to best try to simulate the look you are getting. These console variables determine the set point for colors after color grading, so you can adjust where in the 0 to 1 Linear spectrum the color sets after your Color grading. While not ideal it will be the best solution if you are wanting to stay in 4.9.

Thank You again for your patience -

Eric Ketchum