match-grade
Skill4 starsChanged 10 days agoMITView on GitHub
Colour-grade an edit to match a reference video's look by measuring it — shadow neutrality, highlight warmth, brightness, saturation — and tuning Lumetri until the numbers line up, plus per-shot corrections by kind of light and gentle handling of noisy night shots. Use it whenever the user says the reference "has a warmth/vibe/filter ours doesn't", asks for a cinematic or film look, a specific colour feel, or complains about grain, washed-out or grey footage in an edit.
In this repository6 other files it gives its agents.
- analyze-referenceSkill
- check-renderSkill
- filmcraft-editingSkill
- photo-titleSkill
- recap-editSkill
- select-shotsSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 18
Use frames of the body (not the intro, not black bars or captions):
- Distinctive: line 22
Read it as: shadows R≈G≈B → neutral shadows.
- Very specific: line 33
night: no contrast boost and no shadow lift — they amplify sensor noise.
- Distinctive: line 40
Start from the measurements, e.g.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Matching a grade
Eyeballing a look ("warm", "teal-orange") usually lands in the wrong place — e.g. a teal preset when the reference is actually neutral shadows with creamy yellow highlights. Measure both, compare, adjust, measure again. Two layers: a per-shot correction (balances each clip's light) and one overall look on an adjustment layer above everything.
color_stats.py reports mean RGB in shadows / midtones / highlights, overall brightness and
saturation; with --ref and --ours it prints which way to move.
1. Measure the reference
Use frames of the body (not the intro, not black bars or captions):
python3 ${CLAUDE_SKILL_DIR}/scripts/color_stats.py work/ref --crop 0,112,180,208 --every 6Read it as: shadows R≈G≈B → neutral shadows; highlights with B well below R → warm/creamy; overall mean → exposure; saturation → how punchy.
2. Per-shot correction (by light)
Give each shot a look that fixes its own light before the overall grade (plan look field;
presets in recap-edit's build script):
- haze (grey sea/sky, overcast): +contrast, lower blacks, -highlights, +vibrance, slightly warm
- golden (sunset, sun in frame): -highlights to hold the sun, lift shadows a little
- green (forests, tea, waterfalls): +contrast, +vibrance
- stage / indoor: moderate contrast; indoor sports +exposure
- night: no contrast boost and no shadow lift — they amplify sensor noise; pull blacks down instead,
add
vr_denoise40–50 on the worst shots only. Heavy denoise or a median filter softens the image; users notice blur more than grain. Denoise goes before Lumetri in the effect order.
3. Overall look on an adjustment layer
Start from the measurements, e.g. for "neutral shadows, warm creamy highlights, slightly darker":
{"temperature": 2, "contrast": 18, "exposure": -0.2, "blacks": -16, "whites": 18,
"curve_red": [[0,0],[0.15,0.13],[0.8,0.83],[1,1]],
"curve_blue": [[0,0.01],[0.15,0.17],[0.45,0.45],[1,0.84]], "vignette_amount": -0.7}Curves let you warm highlights while keeping shadows neutral — temperature alone warms everything. Skip film grain on footage that is already noisy.
4. Compare and iterate
Render frames of the edit (export_frames(every: 0.2) or check-render --keep), then:
python3 ${CLAUDE_SKILL_DIR}/scripts/color_stats.py --ref work/ref --ref-crop 0,112,180,208 --ours work/edit_framesApply the advice, re-render frames, compare again. Two or three rounds usually get every zone within a few levels. Then look at the night shots, faces and skies by eye — numbers can match while skin goes orange.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

