How to Match a Color Grade from a Reference Photo
The fastest way to get a cinematic look is to match one you already like. Here's how color matching works and how to do it in under a minute.
The fastest way to get a cinematic look isn't to build one from scratch — it's to find an image that already has the look you want, and match it.
Professional colorists call this "reference grading" and it's how most of the consistent, identifiable looks in film and TV get created. You pick a reference, you match the palette, you refine. It's faster and more reliable than building from intuition.
What color matching actually does
Luttie's color match works in CIE Lab color space — a perceptually uniform space that separates luminance from color. For each channel, it computes the mean and standard deviation of your source image and the reference image, then builds a transfer function that maps one distribution to the other.
In plain terms: it figures out how your image's colors are distributed and how the reference's colors are distributed, then shifts your image's palette toward the reference's. The result is baked into a 3D LUT that you can export and use anywhere.
It's not magic — it won't turn a blown-out outdoor shot into a moody interior. But for images with a similar tonal range and lighting setup, it's remarkably accurate.
What makes a good reference image
Lighting conditions matter more than content. A reference shot in similar light to your source image will match much better than a reference with completely different ambient light.
Good references:
- Screenshots from films with a clear, consistent look (Blade Runner 2049, Parasite, La La Land, etc.)
- Colour-graded photography from photographers whose aesthetic you admire
- Behind-the-scenes or grade examples from colour companies
- Your own previously graded images that you want to replicate
Avoid references with:
- Mixed lighting (tungsten + daylight in the same shot)
- Very limited colour range (all desaturated, or monochromatic)
- Heavy post-processing effects that aren't colour-based (grain, blur, etc.)
How to use color match in Luttie
- Open luttie.app/editor
- Load your source image — JPEG, PNG, RAW, or a frame extracted from video
- In the sidebar, scroll to Color Match and expand the section
- Drop your reference image into the drop zone (JPEG, PNG, WEBP)
- Click Match Colors
- Adjust the Strength slider — 75% is usually a good starting point
The strength slider blends between your original grade and the matched grade. 100% applies the full match; lower values give a subtler transfer.
Refining the match
Color match gives you a starting point — not a final grade. After matching:
- Use Curves to correct any tonal issues (the match may shift midtones unexpectedly)
- Use HSL Secondary to correct any hue shifts on specific colours (skin tones are common)
- Use Colour Wheels to push the shadows or highlights in a different direction
- Adjust Saturation if the match over- or under-saturates
The match gets the overall palette in the right neighbourhood. Your refinements get it exactly where it needs to be.
Exporting the matched grade as a LUT
Once the grade looks right, export it as a .cube LUT file. This lets you apply the matched look to your entire video shoot in DaVinci Resolve, Premiere Pro, or Final Cut Pro — no manual re-grading per clip.
Export single LUT: For one specific strength level.
Export LUT Pack: Exports 50%, 75%, and 100% versions in a zip — useful if you want to leave your editor options when applying the look in the NLE.
Common use cases
Matching across cameras: Shot on two different camera bodies or manufacturers? Color match can help bridge the gap before you grade.
Replicating a film look: Drop in a screenshot from your reference film and match it. Refine for your actual footage. Export as a LUT for the whole project.
Client mood boards: A client sends you a mood board of colour references. Match each one, compare, and pick the closest. Export the winning grade.
Consistency across a shoot: Grade one hero shot, export the LUT, apply across everything else.