On a non ACES shoot, a common practice for building a look with a DP on-set would be to start from a library of 3DLUTs, either built by a company or something I made on a previous project. I found it to be easier to flip through 5-10 looks and start from there, versus trying to discuss using just words
Now doing my first ACES show with Pomfort’s LiveGrade, I’m learning that my 3D LUTs won’t be able to be used.
My question is, has anyone built or used any type of look, such as a film-emulator, that is ACES compatible? I’d love to hear about it.
Hi Ryan,
thank you for your message. I’m sorry for the late response: I actually wanted to reply you earlier in the week, but it has been quite a busy one!
ACES framework allows custom look through the use of an LMT in the chain, which preferably should be applied on a ACES 2065 (LINEAR RGB with AP0) color space. Basically, after your ACES Proxy /CC conversion, you go back to ACES linear and apply a look that, for the very reason that it is applied to the widest ACES gamut with linear gamma, can then be applied to any other ACES output transformation. I’m sure that someone created some sort of film-emulation LUT as LMT, but I’m not sure where you can get them. Maybe Steve Shaw from Light Illusion can create you some if you want.
Anyway, you can try to use your LUTs in ACES, if you convert them for an ACES 2065 space. They will not necessarily work perfectly though, it depends on the LUT characteristics and, of course, their final aim. It will be better to create them from scratch, that’s for sure.
If you want, I can try to convert one for you, so you can test it and let me know what you think.
Although the ideal might be to apply a mathematical transform in ACES 2065-1, since LMTs are often 3D LUTs, those are better applied in a logarithmic space like ACEScc. Evenly spaced 3D meshes do not work well with linear data, so a pre-transform to some form of log or gamma coding is needed.
Because existing LUTs are designed to be applied in a particular colour space (ALEXA LogC Wide Gamut, for example) and the output is often in a different space (BT.1886 video, for example) to get the same visual result from a LUT in an ACES pipeline you would need to concatenate the LUT with transforms from the ACES working space to the LUT’s input space, and from the LUT’s output space through an inverse ODT back to ACES. All the transforms to do this are published in the ACES GitHub repository as CTL files.
While modifying a LUT like this is possible, it is not necessarily optimal. Concatenating multiple transforms can be damaging to the image. As Francesco says, you are probably better off creating a selection of looks from scratch in ACES.
Hi Ryan.
Keep in mind that, as Nick pointed out, LMTs should be applied at a specific part /layer of your ACES pipeline,
It follows from it that a CDL created using an on-set appliance (out of either the same ACES pipeline of later processes, or even out of an ACES workflow) may well not work when used, for example, above the timeline of a color-corrector.
It’s important that the colorspace on direct top of which CDLs were created on-set is colorimetrically equivalent to the one which color-grading and all the finishing is accomplished over (e.g. ACEScc).
Otherwise, the numbers contained in the CDLs are meaningless, as different systems color-communicating with each other via CDL differently interpret those 9-tuples.
It’s like measuring a 9-feet livingroom’s wall, note down just 9, then later buy a 9-meters library for it.
the situation you describe was the main reason behind my effort to create our own software for live grading as Live Grade was pretty useless for this type of workflow in ACES.
There are two important technical issues you need to overcome:
You need to be able to have LMT implemented in LiveGrading application with the CDL applied before the LMT
The same LMT must be compatible with whatever Lab application you are going to use and grading software
I couldn’t find any application that could do either. On top of that, we found it impossible to write DCTL to the spec of LMT that would run in realtime even on a most powerful of Resolve systems at all. We did look into Baselight as well but got a similar response.
To solve the issue I have worked with a great team of developers from France called Firefly in developing our own Live Grading software called Colourlab ( www.colourlab.co )
We have few large productions using it at the moment and we have found the solution to custom LMT’s and their implementation in the post as well as on-set.
You can sign up for a beta release testing on our site and we will be releasing the software to the public in due corse too.