Blackmagic Forum

Have some Ursa Mini 4.6k footage that was shot and somewhere in our workflow we are experiencing a subtle yet very noticeable shift in the contrast/levels between an ungraded ProRes HQ plate and an ungraded plate with VFX rendered out from Nuke as a ProRes HQ. We've traced the issue down to something with Resolve. In Nuke for example, we can toggle back and forth between the two clips (the one from Resolve edit and the one rendered out from the VFX guy) and there is no shift in contrast. When we bring the exact two same clips into the Resolve timeline on my computer (editor) there is a shift. Both are ProRes HQ files. Im not sure where the difference is coming from? Is Resolve reading one clip's IRE range different from the other? The difference looks to be about the same as the difference between a "video" range and the "full" range. Im thinking that has to be it, but then I tried forcing the VFX clip to read as "video" and once as "full" and they were both different from my ungraded ProRes HQ plate. Any suggestions? Is it a bug? We also checked gamma levels etc and everything matches.

This is in 15b5

In Nuke, is there a gamma monitor curve applied to the VFX to put it into TV gamma, from linear, that’s not in the rendered vfx?

First check that you read the original and comped clip in Nuke with exactly the same settings. It sounds obvious, but is not in practice because Nuke tries to read the clip colorspace metadata and defaults to some setting that might not be the one you got when reading in original clip. For example prores4444 clips from Resolve etc tend to default to gamma 1.8 even when in reality they are 2.2, but prores4444 written in Nuke correctly defaults to 2.2 when written as such.

For the absolute ground truth in Nuke you can read both clips in as raw data (use the checkbox in Read node) and see if they still look the same.

Second thing to check again is clip interpretation settings in Resolve.

Hendrik Proosa wrote:First check that you read the original and comped clip in Nuke with exactly the same settings. It sounds obvious, but is not in practice because Nuke tries to read the clip colorspace metadata and defaults to some setting that might not be the one you got when reading in original clip. For example prores4444 clips from Resolve etc tend to default to gamma 1.8 even when in reality they are 2.2, but prores4444 written in Nuke correctly defaults to 2.2 when written as such.

For the absolute ground truth in Nuke you can read both clips in as raw data (use the checkbox in Read node) and see if they still look the same.

Second thing to check again is clip interpretation settings in Resolve.

Yes there is a shift happening in Nuke when we check the raw data and toggle between the VFX shot rendered from Nuke and the original ungraded plate, though we are not sure how to fix that from this point.

Grab the original prores, export a dpx, load in timeline ad plate, go in nuke, do the effects, abs export dpx: they should match.

Keep decoding and encoding prores does not seems a good idea to me

Michael McCaffrey wrote:Yes there is a shift happening in Nuke when we check the raw data and toggle between the VFX shot rendered from Nuke and the original ungraded plate, though we are not sure how to fix that from this point.

So what is your gamma interpretation set to for original and comped version? And what is set in Write node? Post screenshots of whole properties panel knobs for these nodes. In your original post you wrote that you had a match in Nuke, does this still stand and you have a discrepancy only in raw mode? Does the comp result (not rendered, but viewed) match the original input in Nuke? If raw mode shows difference, this should also show it.

And try some other output format as Walter suggested, dpx or exr.