Do Color Transform + Color Transform Invert nodes break Match Grain nodes? I have two setups below, identical besides the Color Transform and getting No Result when I Input Transform from 709 - ACEScg and invert downstream. What’s happening here?
Color Transform Setup
No Color Transform Setup
Haven’t looked in detail. But I do remember from the discussions about the node:
It does do internal color transforms, and it reads the color space information from the upstream nodes. Also, the color space has to match between all the inputs. So it’s conceivable that either you don’t have a perfect match in color space tags, or that the color space information got lost somewhere.
I’d set batch to display color spaces and resolution detail and make sure they’re all the same. Inlcuding bit depth.
Also check the console for any error messages.
@allklier lesson of the week! you are an educator sir. thank you again.
For reference - “All mandatory inputs must have the same colour space.” My input was a Log-to-Lin (jzp) and I adjusted in the Colour Mgmt setting to get the result I was after
Am I being stupid or does the matte input not work? Ie I have a section of my comp I want regrained and I plug the matte in but it makes no difference , it always regrains the entire frame.
I then have to recomp it again after to make it right.
Anyone else?
Not at my system right now, but there may be a button in the node to enable the matte, it may not be automatic by just connecting it.
@Mepstein Works for me. Matte only applies if you use the disperse mode and click it to be active. Then the dispersion happens within the matte selection only.
Still not got the matte function to work , having to recomp afterwards with original grain plate in some cases . but the matte seems to do nothing for me 
Hey @Mepstein
A few things.
If I’m understanding how you’re working, Match Grain implies you are plugging in:
Input 1. Comp
Input 2. Denoised Plate
Input 3. Original Plate
The key here is that it’s expected that the Comp (Input 1) is working on the Denoised plate. Double check this is the case.
When you hook those up, your Match Grain Analyzes your inputs to put the original grain back over the Comp. Because you are comping on the Denoised Plate, you do not want the Match Grain to only affect a Matte area. You want the Match Grain to affect the whole image, because the whole image needs its grain put back.
The matte only comes into play during Dispersion when you want to take a specific region and not use the Original Plate’s grain, but use a dispersed representation of the original grain. This needs (and uses) the Matte input.
Hope that helps.
Yeah I totally get the way it works , it just seems the matte input is totally redundant for the majority of cases when it would surely be a simple function to have it work. That way when you are doing a comp over the Original Noisy plate with a denoised patch it just works , like in the old school Regrain node
The Matte input is not for composition, but to isolate the region affected by the grain dispersion.
Could it be for composition as a Feature Request in the future? Or another matte input for this function?? 
You can submit a General Improvement over at flamefeedback.autodesk.com, but keep in mind that adding composition in the node itself would a level of complexity to it (in the form of multiple additional controls) and I’m not sure this is something we would like to do.
You should not be doing composition with MatchGrain. In fact you should not even be doing any matting within this node, even if only for grain application. See my video I did all about this.
But then you would loose the ability to use this matte input for the grain dispersion at the same if the area is not identical to your comp matte?
Seems like more room for error than just using a simple comp node in your batch. Computationally not anymore load either way.
If anything maybe just rename the matte input ‘Dispersion Matte’ to make its purpose apparent.
One of the thing that has been on my radar for a little while is to create a different type of input named “contraint” or something like this to help differentiate the inputs that receive a Matte for compositing and the ones that receive a matte only for a constrain of the effect, without performing composition. That would apply here.
Unfortunately, this is not really high on my list of priority.
I think this would be a nice cleanup item, but understandably low on the list.
Kind of like mattes in selectives are also not just comp mattes, but also influence the strenght of the effect.
There are plain mattes, and then there is the special stuff.
thought i would give some Kudos where they are due: i’m finally on 2025 and using this node and it’s great. well done @fredwarren and team. love match grain.
Couldn’t agree more, @TimC, and I’m so glad you can finally join in on the fun! They really nailed it with that one. I’m using it on basically every single comp I work on. Huge game changer for most of my day-to-day.
Completely agree. Such a great addition.
Thanks for the good words. We are happy to know you like it.
