Start with size, not colours
The most common question about thread choice is “how many colours do I need?”, and the honest answer depends on how big the design is. A stitch is a single flat patch of colour. To show a gradual change from light to dark, you need enough stitches for the shades to sit next to each other in recognisable bands.
On a design 60 stitches wide, a smooth gradient might be only three or four stitches long. Squeeze ten shades into that space and most appear as single stitches, which read as noise rather than shading. On a 200-stitch design the same ten shades each get a band of their own.
What our test shows
To put numbers on this, we converted the same test picture (a sky, sun, grassy hills and a shaded ball, described in our confetti guide) at three widths, with colour limits from 4 to 64. For each chart we measured how far each stitch’s thread is from the patch of picture it replaces, using the CIEDE2000 colour difference (ΔE) explained below.
More colours help, then stop helping
Average colour error per stitch, by colour limit and design width
- 50 wide
- 100 wide
- 200 wide
Show the numbers
| Colour limit | 50 wide | 100 wide | 200 wide |
|---|---|---|---|
| 4 | 6.7 | 6.7 | 6.7 |
| 8 | 5.0 | 4.8 | 4.6 |
| 12 | 4.4 | 4.7 | 4.4 |
| 16 | 4.3 | 4.2 | 4.2 |
| 24 | 4.2 | 4.1 | 4.1 |
| 32 | 4.1 | 4.1 | 4.1 |
| 48 | 4.1 | 4.1 | 4.1 |
| 64 | 4.0 | 4.0 | 4.0 |
Colour accuracy improves quickly up to about 12 colours and then flattens. From 16 to 64 colours the average error fell by only about 0.2 ΔE, too little to see. What the extra colours do add is cost and clutter:
Past 16 colours, most new colours are rare ones
Colours covering under 1% of the design, by colour limit and design width
- 50 wide
- 100 wide
- 200 wide
Show the numbers
| Colour limit | 50 wide | 100 wide | 200 wide |
|---|---|---|---|
| 4 | 0 | 0 | 0 |
| 8 | 0 | 1 | 0 |
| 12 | 1 | 2 | 1 |
| 16 | 2 | 5 | 4 |
| 24 | 9 | 10 | 11 |
| 32 | 20 | 19 | 15 |
| 48 | 28 | 23 | 22 |
| 64 | 33 | 28 | 30 |
On the 50-stitch design, going from 16 to 64 colours added 31 rare colours, nearly doubled the isolated stitches and dropped the stitchability score from 76 to 36. The 200-stitch design absorbed the same extra colours with a much smaller penalty, because each shade has room for a band.
A sensible starting range
From the test, and from how the generator scores charts, these are good starting points:
| Design width | Simple graphics | Photos | Why |
|---|---|---|---|
| Up to 60 stitches | 4–8 | 8–16 | Past 16, most new colours were rare in our test |
| 60–120 stitches | 6–12 | 16–24 | The generator’s default is 24 colours at 100 wide |
| 120–250 stitches | 8–16 | 24–40 | Room for bands of each shade; score stayed near 80 |
| 250–400 stitches | 10–20 | 30–60 | Detailed portraits and landscapes can use more |
These are ranges, not rules, and they come from one test scene. Portraits with soft skin tones and landscapes with long gradients often reward a few more colours; cartoons and logos need far fewer. The cross stitch pattern generator starts at 100 stitches wide with up to 24 colours, which suits a wide range of photos.
Let the stitchability score guide you
After every change, the generator shows a stitchability score from 0 to 100 with up to two plain-English tips. One factor it weighs is colours for the size: it starts marking a design down above about two colours per 1,000 stitches. If a tip says “a lot for this size”, lower the colour slider or make the pattern bigger. For rare colours, Merge colours used fewer than N times folds tiny amounts into their neighbours so you are not buying a skein for a dozen stitches.
How the colour matching works
Turning a photo into thread colours happens in two steps:
- Find the main colours of the photo. The image is analysed in CIELAB, a colour space arranged so that equal distances roughly match equal differences as people see them. A clustering method (k-means) groups similar pixels and finds the most representative colours, up to the limit you set.
- Snap each one to a real thread. Each representative colour is compared with every thread in the chosen palette, and the nearest wins.
“Nearest” is measured with CIEDE2000. Older formulas measured a straight line between two colours, treating every direction the same. Human vision is not like that: we notice small changes in some hues, such as skin tones and greys, more than in others, and judge saturated colours differently from muted ones. CIEDE2000 adds corrections for lightness, chroma and hue so its numbers line up better with what people see. The smaller the number, the more alike two colours look; a difference of 1 to 2 is about the smallest most people notice side by side. Our methodology page has the details.
Substituting DMC and Anchor
Perhaps you have a chart in DMC codes and a drawer full of Anchor, or a favourite shade is hard to find. The colour key in PDFs from the pattern maker and editor shows the nearest code in the other brand, and the converters at DMC to Anchor and Anchor to DMC look up any code. Anchor numbers are unrelated to DMC numbers, so a conversion always means looking up a match, and as Willow Fabrics puts it, conversions between the two brands are always approximate.
Each suggested swap carries a confidence label based on its CIEDE2000 difference:
| Label | Difference | What it means in practice |
|---|---|---|
| Very close | 3 or less | Very hard to tell apart once stitched; safe almost anywhere |
| Close | over 3, up to 6 | A small difference side by side; usually fine in mixed areas |
| Approximate | over 6, up to 10 | Noticeably different; use with care, away from its original neighbours |
| No good match | over 10 | The other brand has nothing near; keep the original thread |
How safe are swaps overall?
We matched every colour in both palettes against the other brand:
So for most DMC colours a usable Anchor swap exists, but roughly one in four needs care, and a couple of dozen have no fair replacement. These are the DMC colours with the largest gap to their nearest Anchor:
| DMC colour | Nearest Anchor | ΔE |
|---|---|---|
| 680Old Gold Dark | 907Light yellow | 14.5 |
| 3846Turquoise Bright Light | 185Light teal | 14.1 |
| 300Mahogany Vy Dk | 1088Very dark brown | 13.9 |
| 3829Old Gold Vy Dark | 856Brown | 13.3 |
| 918Red-Copper Dark | 944Dark brown | 12.8 |
| 919Red-Copper | 310Dark brown | 11.8 |
| 3844Turquoise Bright Dark | 187Teal | 11.7 |
| 700Green Bright | 229Dark green | 11.7 |
Deep golds, coppers and topaz browns and the bright turquoises are the weak spots, which matters for autumn leaves, metal and wood effects, and tropical water.
When a close match is not close enough
A swap rated “close” can still cause trouble in two situations:
- Large flat areas. A sky or background stitched in one shade shows a substitute clearly if part of it is worked in a different thread. Use one brand for the whole area.
- Neighbouring shades. If a chart uses two very similar colours side by side, substituting one can make them collide into the same colour or spread too far apart. Check that the substitutes keep the same light-to-dark order.
The brands also feel slightly different: DMC has a tighter twist and silkier sheen, Anchor a softer hand. Mixed in one area, that difference can show even when the colours match. When in doubt, lay the skeins on the fabric together in daylight. Your eye is the final judge. Our guide to converting thread colours goes further into swaps and stash shopping.
Practical colour tips
- Check your darkest and lightest threads. Photo patterns depend on the darkest shade for definition. If it has been swapped for something lighter, the whole design can look washed out.
- Think about the fabric colour. A light blue design on cream Aida looks different from the same threads on white. If you skip the background, the fabric becomes part of the palette.
- Strands change the look. Two strands on 14-count give solid coverage; one strand lets more fabric show and makes colours look lighter.
- Group similar colours in the key. The colour view in the generator and editor makes nearly identical shades easy to spot, and the editor can merge two colours or replace one from the full brand chart.
Buying the right amount
Thread use depends on stitch count, fabric count and strands. As a guide, one 8 m DMC skein covers about 2,200 full stitches on 14-count with two strands. The PDF colour key estimates skeins for each colour, and the skein calculator does the same for any single colour. Our guide to how much thread you need shows the working.
The full test results
| Width | Limit | Used | Avg ΔE | Under 1% each | Isolated | Score |
|---|---|---|---|---|---|---|
| 50 | 4 | 4 | 6.69 | 0 | 0.0% | 100 |
| 50 | 8 | 8 | 5.04 | 0 | 0.7% | 94 |
| 50 | 12 | 12 | 4.37 | 1 | 2.5% | 83 |
| 50 | 16 | 15 | 4.29 | 2 | 3.4% | 76 |
| 50 | 24 | 21 | 4.15 | 9 | 4.5% | 58 |
| 50 | 32 | 30 | 4.11 | 20 | 3.6% | 50 |
| 50 | 48 | 38 | 4.06 | 28 | 6.1% | 37 |
| 50 | 64 | 45 | 4.04 | 33 | 6.2% | 36 |
| 100 | 4 | 4 | 6.65 | 0 | 0.0% | 100 |
| 100 | 8 | 8 | 4.79 | 1 | 0.7% | 95 |
| 100 | 12 | 11 | 4.68 | 2 | 1.0% | 92 |
| 100 | 16 | 15 | 4.21 | 5 | 1.3% | 88 |
| 100 | 24 | 22 | 4.11 | 10 | 2.0% | 81 |
| 100 | 32 | 29 | 4.08 | 19 | 2.3% | 75 |
| 100 | 48 | 34 | 4.07 | 23 | 3.1% | 72 |
| 100 | 64 | 40 | 4.05 | 28 | 3.6% | 69 |
| 200 | 4 | 4 | 6.66 | 0 | 0.1% | 100 |
| 200 | 8 | 8 | 4.63 | 0 | 0.7% | 98 |
| 200 | 12 | 11 | 4.36 | 1 | 1.1% | 94 |
| 200 | 16 | 15 | 4.21 | 4 | 1.4% | 89 |
| 200 | 24 | 22 | 4.14 | 11 | 1.8% | 82 |
| 200 | 32 | 27 | 4.12 | 15 | 2.3% | 80 |
| 200 | 48 | 34 | 4.06 | 22 | 2.5% | 78 |
| 200 | 64 | 42 | 4.05 | 30 | 3.3% | 75 |