FilterPictureA PictureEditor.com tool

Turn a photograph into halftone dots

Rendered on your computer, not on a server.

A screen measures your picture cell by cell and draws one mark per cell whose area is the ink that cell asked for. Everything on this page is a consequence of that one rule.

Drag a photograph in, or click to find oneA photograph with real tonal range screens best. Flat graphics and small type have nothing for a screen to measure.

Nothing to hand right now?

Step 1 of 4 — Open

Bring in a picture and you'll see six effects on it at once. Pick the one that looks right, then push it further.

A halftone is a measurement, not a texture

The thing that separates a halftone from a dot pattern laid over a picture is that every mark is derived from the photograph underneath it. The lab lays a grid across your image at whatever angle you choose, averages the pixels inside each cell, and draws a single mark whose area equals that average. Dark cell, large mark. Pale cell, small mark. Stand far enough back and the eye integrates the marks into the tone they were measured from, which is the entire invention.

That is also why the cell average is taken properly rather than by reading one pixel per cell. A single pixel is not the average of the area a dot is standing in, and a screen built that way flickers on fine detail and smears on noise. The lab runs a box average across the picture at the cell’s own radius before it screens, so what each dot reports is genuinely what is underneath it.

The four controls, and what each one changes

Cell size, 2 to 64 pixels
The side of one square of the grid, measured on the preview and multiplied for the export so the result is identical at both sizes. This is the screen ruling by another name: small is fine and detailed, large is coarse and graphic.
Dot shape
Round is the press default and the one that inverts cleanly in the shadows. Square fills its cell exactly at full ink and gives a harder, more digital look. Line produces a ruled screen where thickness rather than diameter carries the tone. Cross is two bars meeting, and holds a surprising amount of detail at coarse cell sizes.
Screen angle, 0 to 90 degrees
45 degrees is traditional for a single plate, because a diagonal lattice is the one the eye is least inclined to read as a grid. At 0 and 90 the marks line up with the pixel rows and the pattern becomes obvious. In four-plate mode this rotates the whole rosette rather than a single screen.
Paper colour
The ground the ink sits on, and the colour anything transparent is composited onto before measuring. It is a decision rather than a default because a screened picture has no half-tones left to hide transparency in.

One plate against four

Single-plate screening measures brightness and spends it entirely on dot area: one ink, one angle, a black-and-white result on whatever paper colour you chose. Four-plate screening splits the picture into cyan, magenta, yellow and black, screens each separately, and prints them over one another.

The angles are not decorative. Four screens sharing one angle would stack into a single coarse pattern with the plates fighting for the same cells. Separating them by roughly thirty degrees turns that fight into a rosette — the small flower of overlapping dots you can see with a loupe on any magazine page — and the rosette is what stops the eye finding a pattern at all. Yellow gets 0 because it is the weakest ink and its pattern is the least visible.

Plate angles in four-colour mode, before your own rotation is added
PlateAngleWhy
Cyan15°Thirty degrees off magenta, thirty off black
Magenta75°The other side of the same spacing
YellowThe weakest ink takes the most visible angle
Black45°The strongest plate takes the least visible one

Getting a halftone to survive a printer

A halftone drawn for a screen and a halftone drawn for paper are not the same file. On screen a dot needs perhaps six pixels across its widest point to look like a dot; on paper, printed at 300 dots per inch, the same six pixels is a smear. The render scale in the export step exists for exactly this: at 2× or 3× the output runs at two or three times the source dimensions and every length you set is multiplied to match, so the dots stay the same size relative to the picture and simply arrive with four or nine times as many pixels describing each edge.

Save it as PNG. A halftone is the most compressible thing a photograph can become — two colours and hard boundaries — and the lossy formats will put a fringe of noise around every single dot, which on a page covered in dots is the whole file.

Questions the screen raises

What is a cell size, and what number should I start from?
The cell is the square of the screen grid that becomes one dot. Set it to 8 and the lab measures your photograph in 8-pixel squares and draws one mark per square whose area equals the ink that square asked for. Eight to twelve reads clearly on screen at a normal viewing size. Under four the screen sits close enough to the pixel grid that the two interfere and you get a second, coarser pattern that is not in the photograph — the control warns you before you commit rather than after.
Why do my dots merge into solid black in the shadows?
Because that is what a real screen does, and it is the only way a halftone reaches a true black. A circle whose area equals the full cell has a radius larger than half the cell, so at high ink the discs overlap their neighbours. Past halfway the mark inverts: instead of a growing black disc the lab draws a shrinking white one on the grid's corners, which is exactly the behaviour of a press plate and is why the deep shadows close up cleanly rather than leaving four white pinpricks that never quite meet.
Four plates look muddy compared to one. What have I set wrong?
Probably nothing. Four-plate screening lays cyan, magenta, yellow and black over one another as transmittances, so the colours multiply rather than add, and a photograph that was bright on a screen comes back with the weight of ink on paper. That is the honest result. If it reads as sludge rather than as print, the usual cause is a cell size too small for the picture's detail, which puts all four rosettes on top of the same fine texture. Go coarser before you go anywhere else.
Can I keep a transparent background through a halftone?
No, and the reason is structural rather than a missing feature. A screened picture has only two states per pixel — ink or paper — and a half-transparent pixel has no state to be in. So the paper colour is a decision the tool asks you to make, and everything transparent is laid onto it before the first measurement is taken. Pick the colour the render is actually going to sit on and it will composite correctly.

Where to go next