As a vfx supervisor with strong color science knowledge (I have designed the color pipeline for more than one studio) and written many specs for the color workflow on feature films… this is a very frustrating thread to read. A major problem with color science is that the terms are poorly defined, many are overloaded, and half the challenge of having a constructive conversation is to be sure you are all speaking the same language. There aren’t many people who understand digital color well, so the majority of info on the web is wrong or confusing. You have the spend a long time to get to the point where color no longer has any mystery to you.
Having said all that. Shoot log always. But if you are viewing log on set while you are shooting and composing your picture, you are doing it horribly wrong. You need to choose a viewing lut and look at your pictures through that.
> half the challenge of having a constructive conversation is to be sure you are all speaking the same language.
Pipeline Render TD here. From my experience, you know how much someone know color science based on the capability to accept such discussion.
Each hardware/software (camera, monitors, DCC) have different bad terms for a same thing. You must isolate that and understand what each term for a particular hardware means.
Color science being complex + terms being badly defined means you _must_ define your terms once for all, at each new project, as soon as you have more than one person handling footage (aka, all the time).
If you don’t do that, you almost always fall into an inconsistency trap.
What resources do you recommend for learning about color science properly?
Dehancer developer here. I second that. I can feel your pain!
I’ve built color management software as well, and I agree, it’s frustrating to see the misunderstandings. The author seems to be having trouble grading S-Log footage, but the noise he’s seeing suggests the footage was shot in 8 bit log, which is unfortunately possible on Sony cameras. S-Log is a joy to work with as long as a) you shoot it 10 bit with a decent bitrate and b) apply the matching log AND gamut conversion. That leaves you with a pretty flat linear image (good) that will behave the same way as properly interpreted footage from any other manufacturer.