How to make a preset from an Instagram photo

Two very different versions of this question exist, and they have different answers. If the photo is yours and you still have the original file, this is easy and exact. If the photo is someone else's, it is an approximation, and the limit is set by physics rather than by effort.

If it is your own photo

You have the pair: the original file on your drive, and the version you posted. That means the edit is measurable. Save the posted version back down from Instagram or from your camera roll, make sure both are the same crop, and compare them pixel by pixel — for every input value in each channel, the output value tells you exactly what the edit did at that brightness.

That is what PresetForge's before/after mode does, and it is free to use. It handles the alignment, bins the pixels, fits monotone curves through the medians and writes the result as an .xmp. It works even if the edit happened somewhere else entirely — in Instagram itself, in VSCO, in Snapseed, on a phone, years ago. The tool that made the edit does not matter; only the two images do.

1. Find the original. If you edited on your phone, the original is often still in your camera roll as a separate file, or recoverable through the editing app's revert function.
2. Download the posted version at the highest resolution available.
3. Crop both to the same frame. This matters — an uncropped original and a cropped post do not correspond pixel for pixel.
4. Run the pair. Read the curve it produces before you trust it: a sensible curve is smooth and monotone. A wild one means the pair did not align.
5. Apply the resulting preset to a different photo and check it behaves.

If it is somebody else's photo

Now there is no original, so there is no measurement — only inference. And the inference has to fight several things Instagram does to the image on the way to your screen.

Recompression

Instagram re-encodes uploads, resizes them to a fixed width, and applies its own sharpening. Fine tonal gradients get quantised, and the subtle shadow colour you are trying to read may be a compression artefact rather than a decision. Work from the largest version you can obtain, and sample large flat areas rather than small ones.

Display transforms

A screenshot taken on a phone in a wide-gamut display mode is not the same colour data as the file. If you must screenshot, do it on a colour-managed desktop browser rather than a phone.

The scene is not the grade

This is the fundamental limit and no tool escapes it. A warm image might be a warm grade, or it might be a photo taken at sunset. Without a reference frame nothing can separate the two. Any single-image analysis is really answering a different question: "what would I have to do to a neutral photo to give it these statistics?"

That answer is genuinely useful as a starting point — it will land the tonal envelope and the broad colour balance close — and genuinely wrong as a finished preset, because it has attributed the subject's colours to the grade. Expect to correct skin by hand afterwards, every time.

Getting closer by hand

Whether you start from an automated draft or from zero, the manual refinement is the same sequence used in grading to match a reference: read where black and white actually sit, match that envelope with the point curve, then the global cast, then the shadow/highlight split, then per-hue work, then finishing.

The specific thing to look for in social-media looks is the lifted toe. A very large share of the aesthetic people are chasing on Instagram is simply blacks that are not black — the darkest tone sitting around 10% rather than 0%, often with a slight blue or green tint. Get that one thing right and you are most of the way there; the rest is saturation distribution.

Sanity-check on a different photo

A preset derived from a single image will be tuned to that image's subject. Apply it immediately to something with different dominant colours. If the look falls apart, the preset absorbed the scene rather than the grade, and the fix is to dial back the Color Mixer adjustments — they are where scene contamination concentrates — while keeping the curve.