One great trick I have that I want to share is neutralgrading, when you get a plate, even when working in aces or on “raw footage” and even outdoors lighting and lenses “tint” the image a bit.
Sometimes also exposure varies between shots that need the same cg assets or mattepainting to be put in.
So what I do on every plate ingest apart from denoising and delensing is to “whitebalance” the shots, pick a neutral and just slightly adjust the multiply/gain of each color channel to neutralize the plate, then after the comp reverse that change. usually just as a cdl grade.
This will instantely get you better cg integration for almost free, and its easier for lighters to stay consistent between shots, its way more likeley that setups can be copied.
you can also use a mcbeth chart for this, but a smart man once said to me that there is sowmthing about it not beign enough patches and that it can lead to issues , but so far I have used them a lot and its been fine.
Now if anyone knows how to do mcBeths in flame … that would be great!
You can even do more elaborate things like reverse-neutralgrade-shotgrade setups in OCIO and flame.
its great and heavily underused , may it help someone.
Great tip @finnjaeger. I had hoped for a macbeth chart matchbox but no go there. Resolve has that option and maybe it lets you export a 3dlut based on that? That would be useful. The standalone software 3D Lut Creator also has a Macbeth chart tool and is a great app for color correction. It has an OFX implementation but I couldn’t get it to work in batch as an OFX plugin.
you also need to be able to reverse it, which you cant do with a 3D lut but its fine with a CDL
I usually do platepreperation in nuke but looking into getting the same results with flame ideally with bfx or something, here is the nuke tool I am using, this + custom OCIO this is the way to go for
longform and features but I have just used the same workflow for commercials.
Resolve can export any grade as 3Dlut .cube file, including color chart results. Just right click on shot thurbnail and select generate 3D lut. Resulted lut file can be used in Flame with color management node, at least it should work. And Flame`s color management node has a swith to invert 3D luts which do great, job most of a time.
Yepp, Flame color management is great. There is special button for step inversion in custom mode. Applying 3d lut can be that step. Of course it`s not invert something that makes image bw, but most of a time it will do what needed just fine
We did some validations asking artists using other product if they are using the Macbeth charts and nobody was. The main reason was that the chart was never the same colour on different cameras and that they trusted their own manual method better anyway.
Wouldn’t this be a good reason that you would want to use a Macbeth chart? To create parity between different cameras for the same scene? Having that in addition to an “inverse” option would not only allow you to quickly match cameras, but to also “apply” A camera’s look to B camera.
I still think it could be a handy tool to add to the box.
to be honest just grey balancing is all I need in 99% of cases, throw up a 18% greycard and multiply RGB to be exactly 0.18 on every channel (average over the whole card because noise) . I made this nuke gizmo for it that does it (pick region either best fit around current exposure or align all to 0.18) basicaly a basic version of what mmcolortarget does
You could pair both of these (same color picking concept, different references “back vs. macbeth”) into a single invertable node. Now that would be a neat tool…
I can’t say that I’ve ever wanted for a full macbeth chart, but almost every day I would love the ability to quickly match a white and black point, which seems to be a bit of a macbeth thing…
In aces, I’ve been using the McBeth chart a lot in nuke. It’s great, It’s help a lot the 3d.
On a job I did a script to convert the matrix back to a matchbox, I can dig it out if somebody wants to have a look.
I would love a machine learning color match node that would either match A plate to B plate, vice versa or meet them somewhere in the middle, seems like this is screaming for machine learning tools to get the artist 95% of the way there. Funny I already submitted this request 7 months ago, FI-02568