How to create a Lightroom preset from a photo

There are two honest routes from "I like how this photo looks" to "I have a preset that does that". The first is to rebuild the look by eye in the Develop module and save it. The second is to measure the difference between the unedited file and the edited one, and let the measurement write the settings. This guide covers both, and is clear about where each one falls down.

Route one: rebuild it by hand and save a preset

This is the route you want when the photo you admire is your own and you still have the original. Open the photo in Lightroom Classic, press D for the Develop module, and work top down.

1. Set the Profile first, at the top of the Basic panel. The profile is the base rendering — Adobe Color, Adobe Landscape, a camera-matching profile, or a third-party one. Everything below it is relative to that choice, so changing it later undoes your tuning.
2. Set white balance with the Temp and Tint sliders, or the eyedropper on something that should be neutral. Get this right before you touch anything else; every colour decision after it is downstream of white balance.
3. Set the tonal range: Exposure, Contrast, then Highlights, Shadows, Whites and Blacks. Hold Option/Alt while dragging Whites and Blacks to see the clipping threshold.
4. Shape the response in the Tone Curve panel. Since the 9.3 redesign the panel opens on the Point Curve, and you switch between the parametric curve and the master, red, green and blue point curves using the small icons at the top right of the panel — there is no longer a "Channel" dropdown.
5. Adjust individual colours in the Color Mixer — Hue, Saturation and Luminance per band — and split-tone the shadows and highlights in Color Grading.
6. Finish with Effects (vignette and grain) and, if the look needs it, Detail.

Now save it. In the Develop module's left panel, click the + on the Presets header and choose Create Preset…. The dialog lists every setting group with a checkbox. This dialog is the whole game: uncheck anything that belongs to this photo rather than to the look.

• Uncheck White Balance unless every photo you will apply it to was shot under identical light. A preset that forces 5400K onto a tungsten interior will look broken.
• Uncheck Exposure for the same reason, unless you are deliberately building a "push two stops" utility preset.
• Uncheck Lens Corrections, Transform, Crop and Spot Removal. These describe one frame, not a look.
• Leave Tone Curve, Color Mixer, Color Grading, Calibration, Effects and Detail checked — that is where the look actually lives.

Route two: measure the edit instead of guessing at it

Hand-rebuilding works when you know what you did. It does not work when you are looking at a finished photo and trying to reverse-engineer it — you are guessing at a transform you cannot see, and each slider you move changes what the next slider appears to do.

If you still have both files — the original and the edited version of the same frame — the edit is not a mystery. It is a function. For every input brightness in each channel there is exactly one output brightness, and you can read that function straight off the pixels by comparing the two images.

That is what PresetForge does in before-and-after mode. It aligns the pair, bins every pixel by its input value, takes the median output for each bin in red, green and blue separately, fits a monotone curve through the result, and writes those curves into the preset. The remaining colour behaviour — per-band hue and saturation shifts, split toning, grain, vignette — is measured the same way and written into the matching panels.

The output is a per-channel point curve rather than a pile of basic-panel sliders, because a curve is what the edit actually is. Exposure, contrast, highlights, shadows, whites and blacks are all special cases of a curve; solving for the curve directly avoids the compounding error you get from trying to express it as six interacting sliders.

When you only have the finished photo

Sometimes there is no original — a film still, a photo you found, a shot from your own archive whose raw file is long gone. You can still get a usable starting point, but understand what you are getting. Without a reference frame, nothing can tell the difference between "the photographer lifted the shadows" and "the scene had soft light". All any tool can do is analyse the finished image's own histogram, colour distribution and tonal statistics, and produce a preset that pushes a neutral photo toward those statistics.

That is genuinely useful — it lands you in the right neighbourhood, and it is far faster than starting from zero. It is not a measurement, and any product that tells you otherwise is selling you something. Treat single-image results as a first draft you then tune.

Test the preset before you trust it

Apply your new preset to five photos that are nothing like the one you built it on: a backlit portrait, a flat overcast landscape, an indoor shot under mixed light, something with a lot of saturated red, and something almost monochrome. Look for blown highlights, plugged shadows, and skin tones going green or magenta. If it survives all five, it is a preset. If it only works on the photo you made it from, it is a memory of one edit.

Keep the anchor count on your point curves modest while testing. Sixteen anchors can describe almost any transform, but they can also describe noise; a curve fitted too tightly to one image's quirks will posterise on the next. If you export from PresetForge, the free tier caps curves at eight anchors and Pro at sixteen for exactly this reason.