Tier: SMT100: Core Principles
Related: SMT200: Systems · SMT300: Research
SMT102: LUTs Are Constraint Systems
A LUT is not a look.
A LUT is a declaration: a commitment to a tone range, a contrast behaviour, a gamut boundary.
Problem
Colourists and DPs treat LUTs as looks: aesthetic choices applied downstream. This is the wrong model.
When you use a Rec709 manufacturer LUT during production and then grade toward a 2383 inspired look, you have been making decisions inside the wrong constraint the entire shoot.
The contrast relationship, the density, the rolloff: none of it matches what you will actually see in the grade.
A DP making exposure decisions against a Rec709 manufacturer LUT is making them against the wrong rolloff. 2383 handles highlights differently. The constraint mismatch does not just affect the grade. It can corrupt the exposure record. You are working blind and calling it a workflow.
Principle
A LUT is not a look. A LUT is a declaration.
It maps a defined input signal to a defined output space. That mapping is a commitment to a tone range, a contrast behaviour, a gamut boundary.
When you choose a LUT, you are choosing what room you are working in.
Everything built inside that room is built to that room’s dimensions. Swap the room after the fact, and nothing fits the same way.
There is a distinction that must exist before any other decision, or else in post-production you will be modeling versus mimicking.
Modeling
- Understanding palette logic
- Understanding colour contrast
- Building a system that produces similar narrative effects
Mimicking
- “Make it look like The Beach Bum”
- YouTube LUTs
- Trying to match a still frame out of context
Modeling is preparing the signal correctly so the transform can do its work. Mimicking is applying the appearance of that transform to a signal that was never ready to receive it.
A good 2383 look done correctly is modeling: the density relationships, the highlight latitude, the shadow structure all exist in the signal before the LUT renders them. 2383 applied at the end of a grade built against the wrong constraint is mimicking. You cannot model what you did not prepare for.
System
In a scene-referred pipeline, there are two distinct LUT roles that must not be conflated:
Display LUT: what you are referencing right now. From camera to a display target: Rec709 (BT.1886) or P3-D65 ST2084. This controls what you see.
Show LUT: the creative look you are committing to. A 2383 print emulation, a custom network look, a studio-supplied target. This controls what you build toward.
These operate at different stages. The error is treating them as the same decision. Using a Rec709 display LUT during production is not a neutral act. It is a display commitment. If your creative intent is 2383, you are looking at the wrong room while building inside it.
The grade is the work done inside the limits that the Show LUT defines. That target has to be known at stage one: during production, during the conversation with the DP on set. The grade is then the discipline of working backward from that target, not decorating toward it.
On deliverable-spec-driven work (a streamer, a studio, a broadcast spec) the constraint is not chosen. It is assigned. The principle holds regardless: working correctly inside an assigned constraint is the same discipline as declaring your own. The error in both cases is the same: treating the constraint as decorative rather than structural.
Digital acquisition gives us enormous colour volume. Wide gamut. Extended dynamic range. More signal than any previous format. The instinct is to preserve all of it and decide later. But a 2383 print emulation LUT has specific expectations about what it receives. It models the tonal and colorimetric response of a specific print stock under specific input conditions. Feed it the wrong input and you do not get the wrong look. You get an undefined result. Something that resembles 2383 but was never built to receive it.
On productions with multiple camera formats, the constraint must generalize. Each input source requires its own input transform to arrive in the same working space. The Show LUT has to hold across all of them. A constraint that works for one camera package and breaks for another is not a constraint system. It is a coincidence.
The Temporal Constraint
A LUT is also a commitment across time.
A vendor-maintained transform is a living algorithm. The vendor wrote it, and the vendor can change it. A well-built in-house LUT is frozen. It is what it is, permanently. Consider a reconform in ten years on a version of the software that does not exist yet. The grade was built with a vendor-supplied input transform. The vendor updated it. The input transform now renders differently. The film has drifted. That LUT you built in-house? Identical to the day it was made.
Senior colourists on long-form work build their input transforms as baked LUTs for exactly this reason. The pipeline needs to be reproducible, auditable, and immune to software drift. An open, documented algorithm maintained under version control can achieve similar stability. The instability is in vendor-maintained, undocumented transforms, not in algorithms per se. A baked LUT also satisfies studio archive requirements cleanly, where an algorithm with external dependencies may not.
The tradeoff is real: a baked LUT is sampled at discrete points. Interpolation between those points introduces error, particularly in extreme highlights and deep shadows.
A well-implemented algorithm running at higher numerical precision is mathematically more accurate. The choice between them is a choice between stability and precision. For archive-grade work, stability is often the correct call. Name the tradeoff. Make it deliberately.
Pitfalls
- Displaying in Rec709 Manufacturer LUT during production while the creative intent is 2383 emulation in the grade: building inside the wrong constraint without knowing it
- Treating the Show LUT as a post decision when it is a production design decision
- Conflating the display LUT technical transform as creative intent
- Adding the look at the end and assuming the grade beneath it will survive the change
- Running multiple competing display LUTs on set: a hero LUT for director and DP, a different reference for the colourist. Competing constraint systems on the same material is a pipeline smell, not a workflow
- Building a constraint that works for one camera format and not others on a multi-camera production
Test
Pull the LUT. Does the grade still hold? If yes, you did not build to the constraint. You built despite it. The grade and the constraint should be inseparable. The work done inside a correctly declared LUT should feel incomplete without it, because the LUT is part of the structure, not paint on top of it.
Distillation
A LUT is a boundary declaration, not an aesthetic choice. You choose it before you work, not after. A Rec709 display manufacturer LUT is not neutral. It is a contract with a specific output space. If your Show LUT is 2383 print emulation inspired, you have been breaking that contract from frame one.
Using the wrong colour pipeline during production while grading towards 2383 emulation in post does not just affect the look. It affects exposure decisions, density relationships, every creative call made on set.
The distinction is modeling versus mimicking. Modeling means the signal was prepared to be received. Mimicking means the appearance was applied to a signal that was not ready. 2383 can only do its work if the input is ready for it.
A baked LUT is also a constraint in time: it does not drift, does not update, does not depend on a vendor. For work that must survive a decade, immutability is a form of craft discipline.
Digital gives us more colour than any previous format. The discipline is in declaring the constraint early and working honestly inside it, whether you chose that constraint or inherited it. Colour that serves the narrative is colour that knew its limits before the camera rolled.
Related Concepts
Principles
SMT101: Exposure Is Structural
SMT105: Robust System Design in Colour