How to Turn a Photo into a Cross-Stitch Pattern

Turning a photo into a cross-stitch chart used to mean printing a grid, holding a pencil and squinting at a photo for an evening. Software does that part now - but most converters simply posterise the image and hand you the result, which is why so many photo-generated charts end up unstitched at the bottom of a project bag.

The fix is to stop treating the conversion as the finished article. A generated chart is a first draft: in about a minute it gets your colours onto threads you can buy and your shapes onto the grid. Then you edit it - the eyes, the outlines, the one background shade that fights with everything else - until it is a chart worth spending forty hours on.

This guide walks through the whole process: picking a photo that will actually work, choosing a size, understanding what the conversion does to your image, editing the draft it hands back, and exporting something you can stitch from.

Step 1: Pick a photo that can survive the grid

A cross-stitch chart is a very low-resolution image. A 150-stitch-wide chart has roughly the detail budget of a 150-pixel-wide thumbnail, in maybe 30 colours. Everything you choose has to survive that reduction.

Photos that convert well share four traits:

• The subject fills the frame. Crop before you upload. A dog's face filling 80% of the frame reads clearly; the same dog standing in a garden becomes a brown smudge.
• The subject separates from the background. Contrast in brightness matters more than contrast in colour - a dark cat on a dark sofa has plenty of colour difference and almost no tonal difference, and the chart will blur the two.
• The light is even and directional. Soft daylight from a window is ideal. Hard flash flattens features; backlighting turns the subject into a silhouette.
• Focus is sharp where it counts. Eyes, nose, lettering. Background blur is fine and often helps.

Photos that fight back: group shots, busy scenes, heavy shadow, motion blur, low-light phone photos full of sensor noise, and anything already compressed to death by a messaging app. If you have the original file, use it rather than a screenshot of it.

There is more detail on choosing a source image - including what to do about busy backgrounds - in choosing the right photo.

Step 2: Decide how big the finished piece should be

Chart size is measured in stitches across, and it is the single decision that most affects both detail and effort. Two numbers follow from it: the finished size in inches (chart width ÷ fabric count) and the number of stitches you will actually sew (width × height).

As rough guidance:

Subject	Stitches wide	On 14-count
Simple logo or icon	80-120	6-8½ in
Pet or single object	140-180	10-13 in
Portrait with facial detail	180-250	13-18 in
Landscape or group	250+	18 in+

Faces need the most stitches. Eyes, nostrils and the corners of a mouth are small features, and below about 150 stitches wide a human face loses the landmarks the brain uses to recognise a person. Pets are more forgiving because fur reads as texture rather than as precise geometry.

Doubling the width quadruples the stitch count - a 200×250 chart is 50,000 stitches, which is a serious project. Start smaller than your ambition suggests. The Aida count and chart size guide covers the fabric-count side of the same decision, and the Aida size calculator does the arithmetic.

Step 3: Understand what the conversion is doing

This is where good and bad tools separate. Converting a photo involves four distinct jobs, and a naive converter only does the first.

1. Downsampling. The photo is averaged down to the chart grid - one colour per stitch. Averaging (rather than point-sampling) matters: it keeps edges smooth instead of jagged.

2. Colour reduction to real thread. The chart must use colours you can buy. That means snapping each stitch to the nearest DMC floss rather than to an arbitrary RGB value. Nearest is also subtler than it sounds - measuring distance in plain RGB produces the classic failures where skin tones go green and blues go purple. Perceptual colour spaces (CIELAB with a ΔE2000 metric) match what your eye calls "close".

3. Cleanup. A raw reduction leaves confetti: isolated single stitches in a colour used nowhere else, and dozens of near-identical shades separated by a hair. Both are miserable to stitch - you thread a needle for one cross, then change. Good cleanup merges perceptually-close threads, removes orphan stitches and smooths noisy gradients while protecting real edges.

4. Backstitch. Cross stitches alone make faces mushy. A thin backstitch outline along the important edges - the line of a jaw, the rim of an eye, the edge of lettering - is what makes a photo chart look intentional rather than pixelated.

If a tool gives you a chart with 60 colours, 200 single stitches and no outlines, it has done step 1 and stopped. And a tool that does all four but hands back a flat picture you cannot change has stopped one step too early: no converter knows which details you care about.

Step 4: Tune, then check the chart honestly

Once you have a preview, judge it on three things.

Colour count. Fewer colours mean less thread to buy, less confusion and a cleaner look; more colours mean smoother gradients. For most photo charts, 20-35 threads is the sweet spot. Below 15 the image flattens; above 45 you are mostly buying skeins you will use for forty stitches.

Squint at it. Step back from the screen, or half-close your eyes. This is the oldest trick in painting and it works perfectly here: at a distance the stitch grid disappears and you see what the finished piece will look like. If the subject does not read from across the room, no amount of extra colour will fix it - go back and crop tighter or increase the size.

Look for confetti. Scan for single stitches floating in a field of another colour. A handful is fine. A rash of them means the cleanup is too weak, or your source photo is noisy.

Step 5: Make it yours

This is the step that separates a chart you stitch from a chart you abandon, and it is the one most converters skip entirely. Software cannot know that the catchlight in an eye matters more than a patch of background, that a logo has to be the client's exact red, or that you want a date in the corner.

In an editor that works stitch by stitch, each of those is a minute's work:

• Fix the face. Darken a pupil, put the catchlight back, straighten a jaw line the grid broke into steps.
• Swap a thread. Replace one DMC shade with another everywhere it appears - useful when two greens in the legend are impossible to tell apart under lamp light, or when you already own 3346 and not 3345.
• Quieten the background. Fill a region with a single colour, or work it in half crosses so it recedes and uses less floss.
• Smooth the staircase. Where two colours meet on a slant, three-quarter stitches cut the corner so the edge runs straight instead of stepping.
• Draw the lines yourself. Automatic backstitch follows the obvious edges; you add the whiskers, the lettering and the signature no edge detector will ever find.
• Clear the confetti that survived. One click each, and the thread list gets shorter.

None of it risks anything: undo goes back dozens of steps, the photo is untouched, and the chart can be regenerated from scratch if an experiment goes wrong.

Step 6: Export something you can stitch from

A finished chart is more than the picture. The PDF you take to the sofa should contain:

• a symbol chart, black on white, with a grid marked every 10 stitches and a bold centre line;
• a colour chart for reference;
• a legend mapping each symbol to a DMC code and name;
• a thread list with stitch counts and the number of skeins to buy;
• finished size for at least the common fabric counts;
• page tiling with overlap, so a large chart prints on A4 or Letter without losing stitches in the gutter.

Print one page and check you can tell the symbols apart at arm's length before committing to the whole thing. Symbols that look distinct on a screen sometimes do not on paper.

A realistic checklist

1. Crop tight on the subject; use the original file, not a screenshot.
2. Pick a width from the table above - round down if unsure.
3. Generate, then reduce the colour count until the palette is 20-35 threads.
4. Squint at the preview. Recrop if the subject does not read.
5. Add backstitch on eyes, jaw and any lettering.
6. Edit: fix the eyes, swap any two threads you cannot tell apart, quieten the background, clear the strays.
7. Export the PDF, print one page, check symbol legibility.

Try it on your own photo

Stitcherry does all four conversion jobs in the browser - perceptual DMC matching, confetti cleanup, automatic backstitch, a print-ready PDF - and then hands the chart to you in a full editor: brush and fill, thread replacement, half and three-quarter stitches, backstitch drawn by hand, and undo behind all of it. Generating, editing and previewing are free - open the editor and drop in a photo.