Why I Built a Custom False Color LUT Builder (and Why You Need One)

If you shoot on Sony, Panasonic, Canon, or any camera with a log profile, you’ve probably hit this wall: How do I accurately monitor exposure without losing the ability to see the actual scene?

Standard false color LUTs paint your entire image in a neon rainbow — great for a quick exposure check, terrible for everything else. You can’t judge white balance, skin tones, or focus when your subject’s face is covered in green and yellow gradients. But relying on a Rec.709 LUT alone is a guessing game for highlights and noise floors.

I wanted a better way. So, I built lut-builder.

The Problem: The “All-or-Nothing” Monitoring Trap

Most cameras don’t have customizable false color built in. To get it, you either buy an external monitor or load a static .cube file. But static LUTs have three fundamental flaws:

They are rigid: You can’t move the exposure bands.
They are distracting: They paint over the entire image.
They kill white balance monitoring: When your subject’s face is covered by an exposure band, you have no idea if your color temperature is off.

This last point is what really broke my workflow. I do a lot of run-and-gun recording — there’s no time to pull up a waveform. I needed to glance at my monitor and immediately know if I’m exposed correctly and if my white balance is drifting. Traditional false color forces you to choose one or the other.

The Solution: Two Modes, One Tool

lut-builder is a free, open-source Python CLI tool that generates professional .cube files. It supports two distinct monitoring modes:

Fill Mode

Covers the full image with exposure-banded false color — similar to what you’d get from a traditional false color LUT or an external monitor. Every stop gets its own color across the entire frame. Great when you just need a fast, full-frame exposure read.

Band (Stops) Mode — The Hybrid Advantage

This is the mode that changes everything. Instead of painting over your whole image, it overlays thin bands of false color only at the exact stops you define, on top of a full Rec.709 base image.

The result: you can see your actual colors, skin tones, and white balance — with just a subtle color marker showing exactly where your exposure sits. You can leave this on your camera monitor all day and still direct, focus, and check color at the same time.

What makes it accurate?

Proper color science: Under the hood, the tool uses the colour-science library — a dedicated, well-tested library for color space conversions. It decodes your log profile to linear light, derives CIE Y luminance using your camera’s specific gamut matrix (S-Gamut3.Cine, V-Gamut, etc.), calculates the stop values from that, then converts back to Rec.709 or Rec.2020. The stops you see are what the sensor actually captured — not a BT.709 approximation.
Tailwind color palette: Pick band colors by name (emerald-500, yellow-400) or enter any hex code. The suggested defaults look clean on a monitor.
Band width control: From razor-thin (±0.25 stops) to wide fill (±2 stops). Thin bands let you see the image; wide bands work like traditional false color.
Crushed blacks & clipped whites markers: Optional indicators at the floor and ceiling of your log profile.
Monochrome base option: Strip the base image to grayscale if you only need exposure data with no color distraction.
IRE mode: For advanced users who prefer to set bands by IRE value instead of stops.
Terminal preview: Before you export, the CLI shows you a visual preview of what the LUT looks like so you can tweak before loading it to your camera.
Save your config: After building, you can save your settings as a reusable config file — reload and modify anytime without starting over.

Getting Started

Installation

You need uv — a fast Python package and version manager. Install it once, and you never have to touch Python manually.

Windows:

Install uv
Download or clone the repo from GitHub
Double-click build.bat

Mac:

Install uv
Download or clone the repo from GitHub
Run build.command

Advanced (any platform):

uv run lut-builder

Building Your LUT

The CLI walks you through every step interactively:

Choose your log profile — S-Log3, V-Log, Canon Log C3, RED, ARRI, and more
Set your display target — Rec.709 (most cameras) or Rec.2020
Choose LUT size — 33 works for most cameras including Sony and Panasonic; use 65 if your camera supports it
Pick your mode — Fill or Band/Stops
Define your exposure bands — e.g., -2, -1, 0, 1, 2 as comma-separated values
Choose colors — accept suggested Tailwind defaults or pick your own
Set band width — razor, standard, or wide
Add optional markers — crushed blacks, clipped whites
Export your .cube file

Decoding Your LUT Later

If you ever forget which color means which stop, just open the .cube file in any text editor (VS Code works great). The tool writes the full band legend as comments at the top of the file — stop values, color names, and band widths are all documented right there.

Loading into DaVinci Resolve

In DaVinci Resolve, go to Color → LUTs panel
Right-click → Open File Location
Drop your .cube file into the folder (no spaces in the filename if you plan to use it on a camera)
Right-click the LUTs panel → Refresh
Find your LUT under Imported LUTs and apply it to a node

From there you can preview exactly how your camera monitor will look before you ever put the LUT on the camera.

Watch It In Action

Built with AI, For the Community

This project was built using Claude Code to bridge the gap between complex color math and a tool anyone can actually use. The colour-science library handles the hard part — proper gamut matrix decoding, linear light transforms, and display conversion — so the output is correct, not just close enough.

I made it open-source because proper exposure monitoring shouldn’t require an expensive external monitor. If you want to contribute or add a profile, pull requests are welcome.

Download or contribute on GitHub: https://github.com/Today20092/lut_builder

Ayoub is a creator focused on merging AI with cinematography tools to streamline the filmmaking process.