Using LUTs with DaVinci Wide Gamut not working.

I realize this must not be a widespread problem, although I’ve seen others on different forums present the same issue :

I’m suspecting that somehow the normalization LUT (chosen from the LUT list either at the node level or in the media pool) is somehow being applied a second time which results in unusable super saturation and increased density of the image.

I am working with Sony sLog3.cine footage which should require a user chosen normalization LUT.

In project settings, if I set up the color management as DaVinci YRGB the clip looks like regular log footage. If I choose YRGB COLOR MANAGED and DaVinci WIDE GAMUT in the color management preset box, The clip is converted to an image close to the original scene minus a bit of contrast and saturation. DEFINITELY not as flat as SONY log footage and definitely in proper range. However it has no LUT applied (that I can find.)

If I choose the Sony LUT for Sony sLog3.cine to rec709 for Venice (or any other LUT for that matter) the image then becomes unusably oversaturated.

I can hand correct and bring into proper range the non-LUT footage. However I am interested in making use of the Sony created LUT FX9 footage to bring it as close to the Sony Venice color science as possible as that is what the LUT was created for.

you are adding two color transform if using an input transform and then another input transform LUT on top

- the input transform in RCM is published by the camera mfg to transform the camera's output into a scene linear working space
- a lut is typicaly doing the same thing, but targets a transformation to a display refered working space

so use a lut (or a color space transform node) for display refered (YRGB)
use an input transform when in working in scene linear (ie ACEScc/cct, RCM)

pick one or the other, but not both

Dermot, thanks for the speedy reply.

To clarify, I am not applying two different LUTs as far as I know.
I’m just seeing that under da Vinci Wide Gamut the behavior looks like a second LUT is being applied somewhere which, if that is what is happening, I can’t track down.

Under DWG in the settings page the log clip is undergoing some kind of conversion. The Sony sLog3.cine clip then appears not completely flat as if it was unmodified, nor is the contrast and saturation in the ballpark as it would be if a LUT has been applied using the old Da Vinci YRGB settings as they were in resolve 16. I understand that if you are working with raw footage resolve will apply the proper conversion LUT based on Meta tags.

However, this is not raw footage. And if resolve’s DWG settings are adding some kind of transform, what is the process for using a different one? Does DWG mean that LUTs applied to a node no longer work anymore?

Try setting up da Vinci YRGB color managed as the science and da Vinci wide gamut as the color management preset on some log footage. Do you find that what you see in the color page is neither log nor nor ball-park corrected footage? Then add a log conversion LUT to a node and see if you don’t get highly oversaturated footage. That is what I am seeing.

While I can hand correct the partially corrected footage, that doesn’t help me much if I want the effect of a particular correction LUT.

Happy to find out that I cluelessly missed something.

Grateful for any help.

In the previous if we hover over LUT we could see the real time difference but on the Version 17.08 Build 33 seems like its broken

Request you to look into it

[Peter Chamberlain]

In the previous if we hover over LUT we could see the real time difference but on the Version 17.08 Build 33 seems like its broken

Request you to look into it

is this a different but to above?
Its working for me on my macOS, maybe you could provide more info.

Peter,
If you are now on this thread, can you comment on my issue? Am I totally off base? Is there a major difference with how DV17 allows us to use LUTs to achieve certain looks when using DWG than in DV16?

Where does one find what LUT/correction DWG is applying to log footage (not to raw that uses camera metadata tags) so that one can choose another without doubling the correction and getting the unnatural
And unusable look I’m getting?

try selecting "bypass" for input transform?
i only have Arri ProRez, RED and ArriRAW on hand, no Sony media
can you send a download link for a sample clip?

Chris:
i tried your set up, selected color managed, and the DWG preset

my media is Arri ProRez / logC, as close i as i can get to your media with what i have on hand

default input with that preset is rec709@2.4
if i change input to Alexa/LogC i see an image normilised for display refered
if i leave the shot in the preset and add the Alexa/ LogC lut in a node i see the same image, a bit darker, and a bit mo' contrast, no biggie tho, seconds to match to input version

the only real world diffrence is the maths of a lut -vs- transform, in that battle transforms are always a better bet

the transforms are published by the camera mfg's

if the transform that is published by Sony does is a mile off matching a lut published by Sony, i'd talk to Sony

Arri have this working reasonably well, maybe try to download a clip from their test media site and see if you see the same resualts i see?

at least that validates the workflow, and takes the discussion away from Resolve, and hands it over to the camera mfg

should note that the same transforms are the basis of both Baselight and Nucoda's versions of RCM, Arri inputs are identical with all three systems

i just do not see much (if any) Sony these days, had a run of F55 for Movies of the Week a few years ago, but mainly see RED or Alexa in the last few years

Dermot-

Thanks so much for your very detailed reply. I am a doc DP after 40 years as a still photographer and installation artist. I was trained as a painter and for a while ran a private color printing lab. So I come by my fascination honestly. In all of my work I have always tried to make “skin speak.“ That’s why my interest in color correction and trying to figure out what assumptions Resolve is making under the hood. It’s a deep plunge.

ARRI and Red are both out of my price range as well as that the Sony FX9 occupies a pretty good sweet spot between good color science, bit depth, DR and technical features like excellent autofocus, stepless auto ND that make it a better choice For a lone producer.

That aside, I’m extremely grateful for the detailed feedback I get on this forum, yours included. As a filmmaker I sometimes feel I am trespassing on the sacred ground of the great color wizards and that more often than not I am simply hunting down my own errors.

Will take your observations and feedback and dive in again to figure out where (most likely) my error is and report back if I find anything useful to anyone else.

Thanks again for all your help.

May I suggest the excellent tutorial from Rippletraining for RCM in version 17?
Your trainer would be Alexis van Hurkman, the main author of the Resolve manual.

Thanks, Uli-
I've been looking into that one and appreciate your recommendation.
Christopher

OK. If you've made it this far, there is an issue and a solution.

First of all, a big thank you to Uli Plank and to Alexis Von Hurkman. After spending $59 dollars (currently discounted from $79) on Alexis' 2.5 hour tutorial on Color management in da Vinci resolve 17, I found my answer.

Color management has changed significantly in Resolve 17 unless one disables color controls which would bring them back in line with resolve 16.

In Resolve 17, under Davinci Wide Gamut controls, Black magic has automated attaching an input color space based on meta-tags supplied by the manufacture in specific footage. If one then tries to apply the normalization LUT that one was using in resolve 16 it results in a completely out of bounds and unmanageable image.

One can choose various manufacturer supplied color space transforms. However, if one wishes to use use a custom LUT to normalize the footage such as Alastair Chapman's Sony slog 3.Cine LUT or the well-regarded
Phantom ARRI LUT, one must disable the input color space and add the LUT of your choice or a color space transform to a node.

I am imagining that most of the professional colorists on this forum are still using resolve 16 workflows and 17 being beta have not come up against this yet.

If this little tidbit of knowledge makes you feel like you're going in the right direction then I would definitely recommend Alexis Van Hurkman's comprehensive tutorial that covers this and many other color workflow
changes new to Resolve 17 that may trip you up if you assume a Resolve 16 workflow.

so use a lut (or a color space transform node) for display refered (YRGB)
use an input transform when in working in scene linear (ie ACEScc/cct, RCM)

If you use a CST in YRGB as the first node to transform the camera gamma/gamut to the timeline (e.g., CIE XYZ/Linear), aren't you now effectively scene linear?