Color grading, automation and what actually works.
Practical articles for colorists, editors and filmmakers: balancing and shot matching, color management in DaVinci Resolve, and an honest look at what AI can and cannot do in a grade.
AI can do real work in a grade: balancing, first-pass matching, isolation. It still cannot decide what a scene is supposed to feel like. Here is where the line sits, and how to test a tool on your own footage.
Resolve 21.1 officially supports AI assistants and has a broad scripting API. An assistant can organize a project, apply LUTs and copy grades. It cannot turn your wheels or press Shot Match. Here is what is documented, what is reported, and what is still experimental.
Automatic matching is good when two shots show the same thing under the same light. It gets unreliable when the frames differ, and most real cuts are like that. Here is why, and how to use matching tools without letting them make the decisions.
AI can apply a look: a LUT, a film emulation, a match to a reference frame. Choosing the look, and making it hold across every scene of a film, is a creative decision. Here is what a look is made of and where automation helps.
A feature film is hundreds or thousands of shots, several sessions with the director and cinematographer, and a set of deliverables. Parts of that can be automated now. The whole thing, unsupervised, cannot. Here is why.
Automation is taking over repetitive technical work in the grade, and that will change the job. The parts that make a colorist worth hiring are judgment, taste, trust and responsibility for the result. Those are much harder to automate.
Automatic correction makes a shot technically acceptable on its own. Grading makes a film look and feel one way, shot after shot, on purpose. They are different jobs, and confusing them is how good footage ends up looking generic.
Viewers rarely notice a good balance. They always notice a bad cut. Balancing and matching are the unglamorous part of grading, and the part that decides whether the look will hold.
Resolve has excellent matching tools: groups, stills, wipes, Shot Match, scopes. Matching still takes a large share of many sessions. The reasons are mostly not about the software.
Most multi-camera mismatches start before any grading: a wrong input transform, a missing setting, a LUT applied twice. Get color management right, then match the parts viewers compare. Here is the order that works.
Both pipelines bring every camera into one wide working space with a log encoding for grading. For balancing, the differences that matter are in the shadows, the output transform and the settings around them. Here is what the official documents say, and what that means at the controls.
Uploading frames of an unreleased film to an online service is a decision with contractual weight. Sometimes it is allowed, sometimes it is not. Here is how to find out, what to ask a vendor, and where offline processing helps.