Choosing a photo that will actually work

A ring of nails and one continuous black thread can do exactly one thing: put more darkness in some places than others. Everything a photograph carries beyond that — colour, sharpness, fine texture, small bright details — is thrown away before the solver even starts.

So the question is not “is this a nice photo”. It is “does this photo survive being reduced to soft grey masses?”

The thirty-second test

Squint at the photo until it blurs, or shrink it on your phone until it is the size of a postage stamp. If you can still tell what it is at that size, it will work. If it dissolves into a uniform grey, no number of lines will rescue it.

That is not a metaphor. The solver works at 500 pixels across and models the thread as covering a fraction of a pixel, which means it is looking at a deliberately blurred version of your image.

What works

• One clear subject filling most of the frame. A face should take up something like two-thirds to four-fifths of the height.
• Front lighting on the subject, with a background that is clearly darker or clearly lighter. A lit face against dark hair and a dark background is close to ideal.
• Real contrast. Not a heavily contrasted “punchy” edit — genuine separation between the light and dark areas.
• A simple silhouette. Pets are excellent for this: a cat’s ears and eyes read at any size.
• Resolution of at least about 800 × 800 pixels after cropping. Below that the solver is guessing at tone.

What fights you

• Busy backgrounds. Every leaf and brick eats lines that should have gone into the face.
• Backlighting. A bright window behind the subject leaves you a silhouette and nothing else.
• Hats, sunglasses, hair across the face. All the landmarks that let a viewer resolve a face at a distance.
• Very dark photographs. Thread can only go on, never come off, so a dark image wants thousands of lines and ends up a solid disc. The studio compresses very dark targets automatically, but it is fixing a problem you can avoid.
• Groups. Two faces in a 50 cm circle gives each one about 20 cm, which is not enough.

Eyes are the hard part

Eyes are where a portrait succeeds or fails, and they are also small, high contrast and full of fine structure — the exact combination the medium is worst at. Two things help:

1. Crop tighter. Making the face bigger in the frame gives the eyes more nails to work with.
2. Turn on “Favour the centre” in the studio’s image settings. It weights the middle of the frame so the solver spends more of its budget there. It is a radial weight rather than face detection, so it only helps if the face is actually in the middle.

Prepare it before you upload

You do not need Photoshop, but a couple of minutes in your phone’s editor pays for itself:

• Crop square, with the subject centred.
• Lift the shadows a little if the dark areas are pure black — detail that is already crushed cannot come back.
• Convert to black and white yourself if you want to see what the solver sees. Or just tick “Show what the solver sees” in the studio, which is the same thing but honest about the gamma and auto-levels being applied.

Test it, do not guess

The whole point of a browser tool is that trying is free. Frame a photo, press solve, and look at the result ten seconds later. If it does not read, change the crop and go again. You will learn more from three attempts than from any advice on this page — including this page.

The expensive part of string art is the four hours of winding, not the solve. Spend the solve budget freely.

Sources

The photographic guidance above comes from people who have made a lot of these: WowStrings on preparing a portrait, vienosiulo on choosing the right photo, and the notes in Automate the Art on how dark backgrounds consume lines.