There’s nothing dik-like about it.
None of us know everything.
The color Management node and the transforms are also related to the color policy of the current flame project.
DM me at phil_man@mac.com or call at 818-746-6131 and I’ll happily show you a color managed project, and all of the required transforms, and then I’ll share it all with you, if you want.
Sometimes I like to hassle Randy because . . . well, just because. So I like him to know when I’m serious and respectful of his help and wisdom, and when I’m just giving him shit to snack on.
The fact that we’re having to converse about different takes and methods and consequences of this indicates a fundamental problem with workflow that should be easy and clickable.
I deserve to be hassled. More often than not.
Create the world’s easiest Color Management Policy
If you want to learn a little about Color Management
The benefits of using ACEScg for Motion Blur…
And here are the benefits of performing different composting tasks NOT in Rec 709.
I’m not a colour management expert by any means, this is just what I do and I hope this is helpful in some way. There’s a lot of ways to skin a cat. We don’t use the rules in the company I’m at. Just a colour management node that will often come in a batch template from one of the lead artists. In a CM node in colour transform mode your goal is pretty simple: get you camera color space from where it’s starting out to ACES and then from aces to ACEScg. I have two colour management nodes here doing the exact same thing one using custom to show what’s happening, the other is an input transform. Both can be inverted losslessly according to my diff matte pulled up to 10,000 gain (at 15000 gain I can start to see little crumblies but that is the most extreme of the extreme and is actually just very slight decimal point discrepancies in 16-bit vs 32bit, when set to 32 bit, no amount of gain on the difference matte can show any grey between the original plate and what I’ve flipped back and forth). And if it can be converted losslessly back and forth and gives me my nice blurs that’s all I really care about unless I’m comping CG).
What flame is doing in the input transform (and what I have reconstructed in the custom node) is taking HD rec709 video, putting that in CIE-XYZ to remove 2.4 gamma encoding, then we need to go into ACES, so we pop the primaries to ACES. And then finally, we need to limit the gamut of the ACES (which is actually ACES-2065) for our VFX uses. That’s where ACEScg comes in, so another primary transform to go from ACES to ACEScg and then we blur with big old bokeh.
When it comes time for colour pipeline big decisions, I tend to defer to the wisdom of the group, but when it’s my shot and I wanna get my nice blurs and I’m in 2d, I generally am taking this route. This may all be wrong of course and I’m always nervous to talk about colour management because it may result in a brutal skewring. But if skewered I must be, than I’ll just know that much more!
It’s very nice when a community shares information.
This is a good forum.
This community shares high quality information.
I once encountered an individual who professed to not only understand and govern the intricacies of color science, for a large firm, but to be an expert.
I poured a fresh cup of coffee.
He explained how to implement an OCIO managed show LUT for purposes of use in nuke, after effects and maya, and specifically how to deal with shot LUTs after source material had been graded by the colorist.
Here’s how that story goes:
A. guess the origination since nobody bothered to collect shoot notes, and the in-house production team did not request such information from the production or camera department.
B. assume that the material is rec709 (flavor not specified), then convert to the assumed camera origination log space using a homegrown LUT in nuke, because nuke is ‘scene-linear’, and ‘floating-point’.
C. assume that this process is reversible, because nuke is ‘scene-linear’, and ‘floating-point’.
D. a show LUT could be added to any viewer in nuke and would correctly transform anything to the correct viewing space.
E. when graded material showed up, assume it was rec709 (flavor not specified) and convert to assumed camera original log space.
F. in this new color space, and viewing through the show LUT, apply a color correction node in nuke and eye match the converted source footage to the converted graded footage. This color correction node (and its inverse) could be considered the shot LUT, and any LUTs provided by the grading department or company could be discarded.
Naturally, by this point I had packed up and ridden home, armed with the full understanding of why this team consistently failed, and why their idea of a pipeline was six one armed children with a rusty bucket…
Great for fairy tales, terrible for a forest fire.
then do this…
This must be the eye matching part… with special glasses…
Interesting. This is exactly what I did, but when I switched between the rec709 and the aces I got a shift, even though my monitor was changing as well. It’s late, I’m tired and I have a(nother) weekend of work in front of me. I’ll check it out again tomorrow when I have a copious free moment. Perhaps there was an error in my ways I didn’t catch.
Nope. You were/are right. Input Transforms involving Rec 709 will include a color shift. But it doesn’t matter. It re-encodes the pixels of the clip such that math works. Then you do your work. Ignore the color change. and invert it at the end. You’lll have an identical clip to your Rec source, but your blurs/motion blurs/linear-friendly operations will be :chefs kiss:.
So this means that while I’m working I’m looking at something that is not my end result? Unless I look at context through the inverted CM? Ignoring the colour change is a big one for me to get over .
It would seem to me the optimum scenerio would be to get my colour as linear/rec709, input transform to ACES, and put a view transform on at the end.
The beauty though, especially when working rec709 post grade, is that you can do what you really need to do in linear (depth of field blur, glows, anything additive, motion blur, etc) and then flip back to working in rec709. It’s totally safely invertable.
For example, I’ll have a batch where I’m keying in log, color graded footage in rec709 as the fill, and my only colour management nodes will be before and after a defocus into blur node for the BG I’m compositing.
Some things simply do not work as well in linear (sharpening/ filtering, frequency separation, keying (aside from IBK) the traditional CC node (mastergrade works better for this)) and some are artist preferences. I don’t like grading in linear. So I’ll CC the BG, flip it into linear, blur it, flip it right back out of linear for it’s comp behind the rec709 FG. That’s why I like having full manual control over the CM nodes and tagging.
Just ask for ACEScg. Or ACEScct.
But you don’t need to. If you’re doing scoop and serve commercials just do Rec and bounce in and out ACEScg or cct as needed.
I kind of do this, but with view transforms. It’s not ideal, but invisible to the naked eye. The benefit is I get to see it in the same way as what I will see when it is all converted back to 709. To paraphrase @GPM, “I’ve never had a client say ‘it looks like you put a view transform on that before you
defocused it.’”
This is exactly the reason that for many use cases, it is not worth it. The divergent opinions about how to deal with this are indicative of the complexity and “why bother” of all of it.
I push my pixels and make pretty pictures, my clients love it, and I don’t have to think about the math or transforms that everyone is talking about.
Don’t get me wrong, I hear you! What gets the job done is what works best and we all work differently and that’s totally cool, everybody does their thing and it’s a happy client that matters at the end of the day. But @randy essentially solved the issue you were initially asking about with a solution that is just two nodes: a colour management node to go into aces and the inverse of it book ended only for the piece of the comp with the lens flare action. To ask about the value of understanding colour management or being able to effectively implement it is a whole different debate. I’ve found it very worthwhile in terms of my compositing. But mileage will vary, and as stated above, all the matters are good results!