That’s partly because the tone curves in DCP profiles are really intended to provide a “look” - they perform operations on linear space and do not encode the output transform. For the most part, most of us (well at least the RawTherapee crowd) make a point of not using whatever tone curve is embedded in the DCP.
Even better - since if you read the white paper in more detail, the camera is not using all code values, normalize the linear light value that corresponds to code value 911. Which is around 16.
I get 46.085537 for code value 1023, and 16.121132 for 911 (might be slightly off since it’s actually 911*4 on a 0-4095 scale…
Using this, I get:
Converted in RawTherapee - no exposure compensation at all, start with Neutral profile, then set the input profile to the DCP @ggbutcher provided, and the output ICC profile to the “s1h” variant generated by the code in the commit above.
VGamut-elle-V4-vlogs1h.icc (8.9 KB)
(preview JPEG extracted from the RW2 @modernhuman posted for reference)
It’s closer than before, but there’s still something going on with respect to saturation, it isn’t quite right… (maybe a black point problem due to lacking proper camconst support?)
And as an “let’s abuse people’s browser color management” experiment, here’s the same JPEG but without the ICC profile stripped:
