Hey, this PR fix the issues with negatives in the AgX Log colorspace.
My initial solution was using a "dummy" LUT, doing nothing, but still clamping between 0 and 1. I actually found it was not working in the pre-log section so I removed it in #10 .
I found another solution using an unintentional behaviour in the CDLTransform (documented in the v2 doc), where negatives are clamped when the power is not 1.
So I just slap it pre-log and post-log and goodby negatives.
It is still hacky because I had to use a small offset in the power value to the 2 forward/inverse transform are not considered a no-op, and the clamp is still applied.
This means there is a small unintended difference in the output data that I don't like to have. But I think we could discard that seeing the benefits it brings.
This works on OCIOv1 and OCIOv2 (behaviour was preserved on v2 for v1 config).
Changes
config.ocio/AgX Log transform chain
added Nuke 13.0.9 launcher to test config on OCIOv1 (all the version above are v2)
Closes #11
Hey, this PR fix the issues with negatives in the
AgX Logcolorspace. My initial solution was using a "dummy" LUT, doing nothing, but still clamping between 0 and 1. I actually found it was not working in the pre-log section so I removed it in #10 .I found another solution using an unintentional behaviour in the
CDLTransform(documented in the v2 doc), where negatives are clamped when the power is not 1.So I just slap it pre-log and post-log and goodby negatives. It is still hacky because I had to use a small offset in the power value to the 2 forward/inverse transform are not considered a no-op, and the clamp is still applied. This means there is a small unintended difference in the output data that I don't like to have. But I think we could discard that seeing the benefits it brings.
This works on OCIOv1 and OCIOv2 (behaviour was preserved on v2 for v1 config).
Changes
config.ocio/AgX Log transform chain