Choosing a frame for a photograph

This page argues against roughly half the shelf for any given picture. That is the useful half of the advice.

Count the detail before you choose

The only reliable test is how much is already happening in the photograph. A market stall, a crowd, a forest, a plate of food: all of these are full of edges, and every one of those edges is competing for the same attention a decorative band wants. A single face against a plain wall, a building against sky, a flat-lay on white paper: these have space around them and can carry a band that does something.

Busy pictures take thin frames. A band of three or four per cent, in one flat colour, does the whole job — it tells the eye where the picture stops and then gets out of the way. Quiet pictures take wide ones, and can survive a pattern, a deep scallop or a glow without the result turning into noise.

The failure mode is easy to describe and hard to notice while you are making it: you widen the band because the frame looks better at that width, and stop looking at the photograph. If you cannot say what the picture is about after the frame goes on, the frame has won an argument it should have lost.

What survives being seen small

Most of these pictures end up in a feed, at a couple of hundred pixels across, on a telephone. That is a brutal filter. A frame whose detail is a small fraction of the band disappears at that size and leaves a slightly muddy line around the picture, which is worse than no frame at all.

Reads small: Drop shadow, floating card, feathered fade, outer glow, a plain scallop at a generous arc width, wide stripes.
Needs room: Terrazzo, herringbone, fine gingham, postage perforation, a pinked zigzag at a small tooth width, torn paper at high roughness.
Depends on colour: Neon rim and inner glow. Against a dark photograph they hold; against a pale one they vanish into the picture.

If the destination is a print or a screen at full size, the second row opens up. A terrazzo band at 300 pixels wide on an A4 sheet is a genuinely different object from the same band at twelve pixels in a timeline.

Colour, in two moves

Every frame here has one to three colour slots, and there are only two decisions worth making with them. The first: pull the band colour out of the photograph rather than off a colour wheel. A muted tone that already appears somewhere in the picture will sit down quietly; a colour chosen independently competes, however pleasant it is on its own.

The second: decide whether the band should be lighter or darker than the edge of the picture. Lighter reads as paper and pushes the photograph forward. Darker reads as a mount and makes the photograph feel recessed. Both are fine. Matching the brightness of the edge is the one option that is not, because the boundary then disappears and the frame stops doing its only structural job.

A note on the depth frames

Drop shadow and floating card are the two frames that work on nearly everything, and it is because they add almost no decoration at all — they add separation. If a picture has to sit on an unknown background and simply needs to stop dissolving into it, start there and stop there.

When to use none

There are three cases where the right answer is to close this page. A photograph that already has a strong graphic edge — a silhouette, a hard horizon, a subject cropped tight against the frame — is doing its own framing, and a band will fight it. A picture going into a layout that already has rules, gutters and margins does not need a second set. And anything destined for a printed piece where somebody else controls the white space should arrive plain, because two margins is a mistake that is expensive to discover at the printer.

Nothing on this site improves a photograph. A frame changes how it is presented and it can make a weak picture look considered, but it cannot make it a better picture, and a wide ornate band on a poor image draws attention to both.
All of this advice is about screens. The colours here are sRGB and no profile is attached on the way out, so a band matched carefully to a photograph on your monitor may not hold that relationship on somebody else's, and certainly not in a lab print.