I made a tonemapper compositing build from code parts i found in the net.Since AgX and Filmic was there as Tonemapper in Blender, i wanted always to try out if its possible to make a tonemapper with Blender nodes.
In the past weeks i have posted already some testrender results from the nodebuild in the AgX threads.And because i dont want to clutter the AgX topic i just started this thread.
The Tonemapper has basicly just 3 functions.The log2shaper,Gamut compression and a Sigmoid contrast shaper.
Today i just want to show some tests i made today .I adjusted the pre log2shaper to match the middle grey of a Arri testimage,to get a EV0 setup.This means if i plug a image to the tonemapper it shows the same middlegrey at EV 0 in Blender.
Then i loaded the Arri bluebar exr test image with this setup.For comparison the image on the topleft is a screenshot from Arritool from the loaded reference file.On the right the tonemapper build with the blue bar exr at EV 0 no adjustment.
Left down is AgX with no look.
I rebuilded a code that converting LogC Arri Wide gamut to Linear Rec709.
It would be even better if all Colorspace transforms would be available with OCIO.Maybe its already doable,however here a screenshot from the CST node.
Left Arri reference screenshot,right LogC AWG into CST node to my tonemapper.
I found a simple way to keep the original Hue intact without rotation.This works for most images,but not for out of gamut images since these have altered hues from the start.
Here a screen with hue preserve.No hue rotation at all color values.
Another test.
I connected the LogC_ARRI wide gamut image direct with the tonemapper,without any CST in between.At the end of the tonemapper i placed a CST node (LogC_ARRI wide gamut to Rec709).
The reason was to see if the CST does a transform without any out of gamut transform.
Here a screenshot on the left ARRI reference tool,right tonemapper with (logC_AWGto rec709)
The difference here is that the ARRI tool converts with AWG3 LogC3 D65 white to Rec709 with a BT.1886 curve and D65 white (image on the left)
I guess with the exact transform it would look identical even with this tonemapper.
The good thing of this test is,it shows how the correct chosen CSTs dont need a gamut compression.(Even this CST code used are not the exact transform workflow as in the Arri tool)
Based on this video i seached for the filter described there.
Then i found paper about the transmission values of all CC Filters for each wavelengths
In the video was meantioned that the Filmmaker used a 81 EF Filter with a green Filter
I have tested both Filter and come to the conclusion that the Filmmaker was using a cyan Filter instead of green.
Here a screenshot from my testing.Have not played with contrast or something just the two Filter.81 EF and cyan CC 50
In the Ring look video,the researcher described the film development process.
The film was filmed with +1/3 stop overexposed.And during development underexposed by 1 stop.
He further mentioned that the filmmaker used a custom made green filter with the half filterstrength.The half green was used in closeup shots for keeping skintones less tinted.
For my testings of this half green filter i used the values from cc20 and cc30 green and its average to get a “custom” half green (with the assumption that full green or cyan has a density of 50)
I guess that for indoor they dont used the 81 EF filter for get more light and maybe just need the 81 EF for balance the outdoor lighting.
Here a test with a custom 25cc green filter only.exposure set to -1.6stops.(-1stop for simulate the development underexposure and -0.6 for the 81 EF filter taken out.
I tryed a new idea to get analog film like saturation with grading
For color reference i used a colorchart which represent Arri Reveal colors to adjust the exr image to get roughly the same swatches.
I loaded a reference ArriWG4LogC4 image into the tonemapper build and created a colorspace transform script for the ocio.config
The howto is here
Since this is just a colorspace transform from Arri wide gamut 4 to the new colorspace (in this case bt-2020 for further tonemapper calc) the LogC4 to Linear transformation has to be made to.
I rebuilded the Log4 to Linear code from the Arri page.
I adjusted the WB with 50% 7300K (with a custom colortemp node group) and reduced the contrast (with a custom sigmoid node group which has a fixed greypoint) to fit the greyscale from the chart
More Alexa35 footage,
Anamorphic Arri WG4 LogC4 to rec709 with the tonemapper.Contrast settings the same as from the Helen and John render.Exposure,WB,saturation density,and gamut threshold.
Instead of converting the Arri AWG4LogC4 to bt-2020 in the tonemapper,i converted it directly to rec709 in tonemapper.
This way i get closer to the original rec709 reference image
Left original rec709,right Arri AWG4LogC4 to rec709,no saturation adjustment,only exposure and gamut threshold.
