high dynamic range loses info

[frueter]

I just had a quick play with this comparing it to a very troublesome retime I had on my last shot. It's looking great, thank you so much for sharing!
I just noticed that in overexposed areas (e.g. white striped shirt in direct sun light) the results were clamped at 1.30831.

Easy workaround is to stop the plate down for retime which solved that issue.

I assume this is a limitation with the model?!
Great to have this in the toolbox!

[nicocronau]

I second this issue. It isn't a limitation to the model, but an issue with the colorspace conversion inside the gizmo.

The gizmo uses an OCIOColorSpace Node to convert Input from linear to sRGB (and vice versa at the end of the gizmo's nodetree). Somehow the OCIOColorSpace sRGB conversion seems to be non-floating.

If you want to stick to the OCIOColorSpace Conversion node, you should select sRGBf (f=float) for the input and output conversion. Otherwise simply use the nuke Colorspace Node and convert from linear to sRGB. There's no visual difference between those two options.

Here is a quick fix:

Close Nuke
-> Go to your RIFE Cattery Folder
-> Locate RIFE.gizmo
(-> backup file, just in case)
-> Edit with editor/notepad++/whatever
-> Find "OCIOColorSpace" (should occur 2 times)
-> in the first occurence change "out_colorspace color_picking" to "out_colorspace sRGBf"
-> in the second occurence change "in_colorspace color_picking" to "in_colorspace sRGBf"
-> save file and restart nuke

Workaround tested in Nuke 15.0

[rafaelperez]

Thanks for reporting this @frueter and thanks to @manwhosoldthew for providing the solution.
I've applied and tested the fix on my end and it looks good.
I just released a new version that includes this fix.