Looking for help from my color science nerds out there. I am working with a feature graded in DCI-P3. I will be doing some additional trailer grading on top of the feature grade. My final deliverables will all be rec709. When using input transform from DCI-P3 to rec709 on the timeline, I get a pretty big color shift (more saturation, darker). However, when I use an input transform that goes from rec709 to rec709 (yes, that’s correct), the image looks identical—no color shift whatsoever. I’m going to go with it for now, but I am pretty confused. I know they have different gammas (2.6 for DCI-P3 and 2.4 for rec709) so I’m not exactly sure why what I did works…anyone know? Should I just keep my mouth shut and not jinx it?
Just as an experiment.
Call it a sanity check because rec to rec makes no sense. I guess that was why you put the question out there. Sorry I have no experience of this particular transform
Shouldn’t it look different, though? I think the goal of the conversion is to make it look as close to the same as possible on a calibrated rec709 monitor as it does on a calibrated P3 monitor. Doing so should change what the picture looks like on your artist monitor.
Ideally in my current case, they would not look different, as the director is quite attached to the P3 grade and wants the rec deliverables to look the same. My monitor supports both P3 and rec709 but I’m not using an viewing transforms on the media, only input transforms.
In theory any footage tagged as P3 will convert to your monitor settings and look correct but turn off your viewing LUTs by selecting bypass and see what is really going on.
With the viewing LUT, yes you would expect them to look the same (ideally) but this can also fool you so I go for bypass as well.
Like I said this isn’t a transform I have done but I dance extensively around LogC, ACES and rec.709
and I always go for a sanity check by using bypass so that I can see the pixels changing and then pull my viewing LUT back up and see them looking identical.
Sorry, I meant that in order to have deliverables that look the same on rec709 and p3 external monitors, you will have to generate deliverables that look different out in the OS, as rec709 and p3 primaries and gamma are different. Flame’s color management might try to hide this from you, but I can’t say for sure because I hand-manage everything.
I’m bypassing everything which is what I meant when I said I’m not using any viewing transforms, only the input transform, so you can see my confusion with there being no discernable difference.
Perhaps I’m missing something here but to me it sounds like you should be using a view transform from source to rec709 as opposed to an input transform as essentially you’re wanting apply said viewing LUT so you don’t see any shift which should be fine as you’re creating your final deliverable so you can bake-in the viewing LUT.
A view transform from DCI-P3 (D65 white) to rec709 video also results in no color shift from the DCI-P3 plates so yes, this is what I have been using. I guess I was more curious about some of the unexpected behavior I was experiencing.
That is why I like using the more manual Colour Transform (custom). You can flick certain bits, like the primaries, on or off, to see the shift.
The view transform is much easier and more automatic but on the right hand side of the buttons you can see the transforms it is using.
It is using a different intermediary, XYZ colour and the results when compared to my custom Colour Transform, are similar but not identical.
If I was going to commit to one or the other I would probably trust the Autodesk View Transform a little more. Especially since I was just making up the ACES one.
If you use rec to rec: There should be no difference, because this input transform does not change anything. Only if Input CS and Working CS are different, a transform is applied.
In my opinion you will always see a difference when you convert DCI-P3 to rec709. Less gamut and a different gamma.
If you now put a DCI-P3 calibrated monitor next to a rec709 monitor and show the matching footage in the respective color space, then it should “look very similar” but not identical, because one monitor can “only” display rec709 colors and nothing more.
This is an interesting topic. Let me share my opinion:
This conversion should be seen as a three step process (although mathematically these steps do not come one after the other in the listed order):
Just a conversion from <DCI-P3, Gamma 2.6, DCI white point> space to <Rec709, Gamma 2.4, D65 white point> space won’t cause any change in colors, except for the out of gamut colors (which Rec709 display can’t do, but are part of the DCI-P3 space).
If after conversion, the content will be evaluated in a dim room with 5 to 10nits ambient light level (and ambient light color temperature of 6500K), it is suggested that a chromatic adaptation transform be also applied with the above conversion.
In a Rec709 viewing environment (with 5 to 10nits ambient light level with color temp of 6500K), the target peak luminance would be 100nits (instead of 48nits, under which the DCI-P3 content would be graded). Here, a viewing conditions adjustment transform should be applied as well with the above conversion. For instance, CIECAM02 based viewing conditions adjustment can be applied.
I should highlight that as of now there is no perfect model for viewing conditions based adjustment. CIECAM02 works reasonably well, but sometimes colorists like to create a LUT rather, which lowers gamma (as less contrast is needed in a brighter environment for perceptual match) and desaturates the image (as perceived colorfulness increases with the luminance - Hunt effect), so as to get a tuned perceptual match after the DCI-P3 to Rec709 conversion (for the Rec709 viewing environment).
The idea behind 2. and 3. is to make the content look perceptually very similar when it is evaluated in a different viewing environment with a different calibration target.
It should be noted that 2. and 3. do not apply, if the Rec709 converted content is to be evaluated in a projection setup in a completely dark room.
Well, this is how we provide Calibration 3DLUTs to our clients, so as to directly calibrate their display (in a Rec709 viewing environment), for presentation of DCI-P3 graded material (or DCI XYZ material, that was graded on a projector in a dark room). Such 3DLUTs include the combined color processing of 1., 2. and 3., and do the required color correction as well for the display.
