For most of the twentieth century, the record of an artwork was a transparency in a drawer. It was made once, by someone paid to make it, and everyone who wrote about the work, insured it, or reproduced it in a catalog worked from that single object. The record was singular, and it was slow. Now the record is whatever image happens to be circulating, and it was probably made on a phone.

Two Different Jobs, One Photograph

A photograph of an artwork gets asked to do two incompatible things. One is documentation: an accurate account of what the object is, for archives, insurance, condition reports, scholarship, and the artist's own record of their practice. The other is promotion: an image built to make someone stop scrolling, click, visit, or buy.

Documentation wants fidelity. Promotion wants impact. A single file is usually asked to be both, and the compromises get made silently, by whoever happens to be editing.

How Institutions Decide an Image Is Good Enough

The institutions doing this at volume have taken the judgment out of it. Under the third edition of the technical guidelines for digitizing cultural heritage materials, published in 2023 by the Library of Congress-sponsored Federal Agencies Digital Guidelines Initiative, a digitized object earns a rating of one to four stars based on measured analysis of the image against defined parameters, not on how good it looks to whoever shot it. The current edition states its color accuracy tolerances in Lab values, which is to say in numbers a target and a piece of software can check between them.

Nobody expects an artist photographing work in a spare room to run conformance software. The transferable idea is smaller and nearly free: put a color reference in the frame, once, and keep the unedited capture. A gray card and a ruler in the corner of a raw file cost nothing and convert an opinion into a comparison.

What the Camera Gets Wrong About Paint

Even a careful photograph misrepresents a painting in predictable ways. Impasto flattens: the physical relief that reads as gesture in the room becomes texture on a screen at best, noise at worst. Gloss varnish resolves into specular hotspots. Deep blacks in oil sit outside what a display can show, so they compress toward one another and the shadow detail vanishes.

Color is the worst of it, and not only because of white balance. Certain pigments fall outside the range a standard screen can reproduce, so the file substitutes the nearest available color. That substitution isn't an error anybody made. It's a limit of the medium the reproduction lives in, and it means a viewer looking at a technically accurate photograph of a saturated cadmium red has still not seen the red.

The Line Between Correction and Improvement

Some editing is straightforwardly corrective. Squaring a canvas shot slightly off-axis, neutralizing the cast from the studio's overhead lights, cropping out the wall, removing the photographer's reflection in the glazing: none of that alters what a viewer understands the work to be.

Other edits do. Deepening the blacks so the composition reads more dramatically. Raising saturation because the piece looked flat on a phone. Erasing a stray canvas staple, a scuff, an area the artist deliberately left rougher than the rest. Replacing the wall behind a sculpture with a cleaner one.

The test isn't whether an edit is visible. It's whether someone who came to see the work would feel they'd been told something untrue. A collector who buys from an image and unpacks something duller than the file won't describe the difference as color management.

The In-House Shift

Galleries that once paid a photographer for every work now shoot in-house, because whoever runs the social accounts can open an AI photo editor like Pixelcut and have a clean image before lunch. Artists without representation have gone further still: for many, the phone in the studio is the only camera the work will ever meet.

This is a real democratization and it deserves saying plainly. Documentation used to be a cost that excluded people, and it isn't any more.

What it also means is that the person deciding how a work gets represented is now, almost always, the person with an interest in how it's received.

The Collector Who Sees It in the Room

Dealers describe the same failure over and over. A work performs well online, someone travels to see it, and the encounter is a small disappointment nobody names out loud. The sale doesn't close and no reason is offered, because a buyer can't easily articulate that the image promised more contrast than the paint delivers.

Images Outlive Their Captions

The image has a long life after the exhibition. It gets reposted, scraped, printed in a listing, folded into a training set, attached to a résumé, and eventually used in an obituary. Somewhere along that chain the caption falls off, and what survives is a file with no record of what was adjusted or by whom.

Keeping the original capture beside the edited version is the cheapest provenance available to anyone working now. It costs storage. It answers questions that would otherwise be unanswerable in twenty years, including the question of what the work looked like before anyone improved it.

Photograph the Work, Not the Sale

The workable discipline is to make the accurate file first, derive the persuasive file from it second, and keep both. Documentation and promotion stop fighting each other the moment you accept they are two outputs rather than one compromise. An artist who works that way ends up with an archive. An artist who edits in place ends up with a marketing asset and no record of what it came from.