Build an Original Repeat Pattern for a Commercial's Designed World
Build an Original Repeat Pattern for a Commercial's Designed World
A repeat fails in two ways, and only one of them looks like a mistake.
The first is the seam. Two tiles meet and something doesn't line up. You can point at it, and you can fix it.
The second is quieter. The tile is technically perfect. The join is invisible, the pattern repeats cleanly, and the wall still looks wrong: a row of dots running dead straight across it, a cluster of leaves all crowding the same corner, a rhythm that keeps tapping the eye while somebody is trying to deliver a line in front of it.
Both come from the same misunderstanding. A repeat pattern is not a picture of a motif. It is the relationship between one tile and its neighbours. A beautiful leaf can make an unusable repeat. An ordinary leaf, arranged with its copies in mind, can carry an entire designed world.
So the work runs in four moves: decide what the surface is doing in the scene, build the tile so its edges belong to the tiles next door, look at a field of many tiles rather than the one, and check the whole thing at the size it will actually be seen. Each move answers a question the previous one cannot.
Choose what the surface contributes to the scene
Before you draw anything, name the surface and give it a job. A patterned wall behind a counter, a fabric on a banquette, a wrap on a package: each one sits at a different distance from the story, and that distance decides how much detail you can afford.
Say the proposal is a café interior. A wall behind the counter is going to be visible in several shots, partly cropped, sometimes behind a small product. Stipulate the wall at roughly 3.2 metres wide and 2.7 metres high, seen from about four metres. Those numbers are invented for this example, but you should invent your own before you draw, because they are what makes the rest of the decisions checkable.
Now separate two things people usually merge: the motif and the tile. The motif is your leaf, your chevron, your mark — the thing whose shape you care about. The tile is the rectangle you will repeat. They are not the same object and they are not the same decision. A small motif in a generous tile gives you air. The same motif in a tight tile gives you texture. Same drawing, different world.
Also decide orientation. Is the repeat upright on a wall, or does it run with a fabric's grain, or wrap around a package where the surface curves away from the viewer? A pattern that reads beautifully flat on screen can slide into nonsense on a cylinder.
Keep the motif original. Copying a protected pattern is a rights problem before it is a design problem, and a treatment is not the place to discover that.
Make the boundary crossings belong to neighbouring tiles
Here is the mechanism. Set your tile at a fixed size — 200 units square, say, because round numbers make the arithmetic visible. Put your motifs in it. Then check every single element that touches an edge, because whatever hangs off the right side has to exist somewhere on the left.
Work an example. Your leaf is 24 units wide and 40 tall. Place one, call it Leaf A, centred at (40, 40). It occupies x 28 to 52 and y 20 to 60. Comfortably inside; nothing to do.
Place Leaf B centred at (194, 140). It spans x 182 to 206 and y 120 to 160. Six units of it hang past the right edge of the tile, which sits at x = 200.
That six-unit overhang is the whole problem. The tile next door, sitting at x 200 to 400, should contain the missing sliver entering from its left edge. So you place a copy of Leaf B translated left by exactly one tile width: 200 units. The copy lands centred at (−6, 140), spanning x −18 to 6. Inside your tile, only the strip from x 0 to 6 shows. That strip is the continuation. It is not decoration; it is the same leaf being cut by a boundary you chose.
Corners need more than edges. If a motif overlaps the top-right corner point, you need three copies, not one: shifted left by one tile width, down by one tile height, and both. Get those three in and the corner neighbourhoods of four tiles all resolve at once.
Two habits make this survivable. First, place the copies yourself rather than hoping the application wraps them for you. Adobe's Illustrator help documentation describes the mechanical route — selected original artwork becomes a pattern definition that you apply from a swatch — and that step is genuinely useful and fast. It also says nothing about whether the join is good, and it shouldn't; the seam is your decision, not the command's. Placing the continuation as a real object keeps the construction visible and keeps the file portable to another application, a colleague, or a printer.
Second, set the tile size deliberately. If the tile becomes whatever bounding box your selection happens to produce, a stray point or a forgotten object silently changes your repeat. A rectangle carrying the intended tile dimensions, sitting behind the artwork, keeps the size from drifting. Illustrator's own page was checked on 18 September 2026 for this piece, and its scope stops at definition and application; nothing in it establishes continuity, rhythm or manufacturing suitability. You are supplying all of that.
Keep the artwork live. A flattened bitmap tile is a dead end the first time the surface scale changes — and it will change.
Inspect the field, not just the seam
Copy the tile until you have a field big enough to judge: several repeats in both directions, enough that the eye stops reading the tile and starts reading the wall. Then look for four different things, because they have four different causes.
Boundary discontinuities are the ones you already know about — a leaf sliced off with no counterpart, a corner where three elements collide.
Accidental stripes are the interesting ones. This is where the repair from the last section can be flawless and the repeat still fails.
Take the tile above and add two dots, 10 units across. Dot C centred at (60, 90), Dot D centred at (160, 90). Both sit at the same height, 100 units apart horizontally. Inside the tile that looks fine — two dots floating in the space between the leaves.
Now tile it. The dots land at x 60, 160, 260, 360, 460 and so on, all at y 90. A dotted rule now runs straight across the wall, unbroken, at exactly the height where the eye wants to rest. You never drew a stripe. The tile made one for you.
Here is the part worth remembering, and it is why "just nudge something" is bad advice.
Say you move both dots down by 40 units. Dot C goes to (60, 130) and Dot D to (160, 130), and the rule is still there. It has simply moved. A horizontal line produced by one family of elements sharing a y-coordinate cannot be removed by sliding that family around, because every copy slides with it. Set the pair back where it started and change a relationship instead: give the two dots different heights, so the line breaks into a zigzag.
So move Dot D from (160, 90) to (160, 130), and leave Dot C at (60, 90). Dot D stays inside the tile — the move doesn't open a new boundary crossing, which you should always check — and the field now has dots at y 90 in some columns and y 130 in others. The unbroken rule becomes a stitch. That is a placement change, and it is targeted: it alters exactly one relationship and leaves everything else standing.
A tile-dimension change is the blunt instrument. Making the tile 260 units tall instead of 200 changes how far apart every repeated row sits, which addresses a genuine problem — a field that is too crowded or too thin. It does not fix a single-family alignment, because that family still shares a height. And it quietly invalidates all your boundary work: every continuation copy you placed sits exactly one tile width or height away, and if the tile gets wider, those copies are now in the wrong spot by the difference. Change the tile and you redo the seams. Know which problem you actually have before you reach for the tool.
The last two things to look for are conspicuous clusters and empty patches. Both are density problems, and both are usually visible only in the field. Resist the urge to declare a pass or fail. Whether a cluster is too much is a judgment about this café, this camera, this product — not an objective threshold, and not something a rule can settle for you.
Compare the repeat at the surface's working scale
Now put the checked pattern on the surface, at two sizes that could plausibly be specified, and look at both.
Using the invented café wall: at one tile = 20 centimetres, the wall carries 16 tiles across and the leaf measures 2.4 by 4 centimetres. That is a sprig, not a motif; the wall reads as texture. At one tile = 60 centimetres, the wall carries about 5.3 tiles across and the leaf measures 7.2 by 12 centimetres. Now it is unmistakably a motif.
The two scales do different work. At the fine scale the wall stays quiet and the product in front of it reads clearly, but the dot lanes land every 20 centimetres, roughly thirteen of them across the wall's height, close enough together that the eye may start reading them as a woven band, or as a wobble on camera. At the coarse scale the lanes fall every 60 centimetres and simply stop being a rhythm problem at all — a handful of rows read as intended placement. The problem you fixed in the tile matters more at one scale than the other. That is not a paradox. It is why you check both before and after.
Notice what the coarse scale costs you. Leaves twelve centimetres tall on a wall immediately behind a small cup will fight the cup for attention. Scale is not only about legibility; it is about who the shot is for.
The wall's width rarely divides evenly by the tile. At 16 tiles across, the fine scale nearly lands; at 5.3, the coarse scale does not. For a dressed set that is often irrelevant, because the camera frame crops inside the wall and the wall's own edge never appears. But the phase still matters — where the crop falls relative to the tile determines what the frame actually shows. If it matters, shift the pattern within the wall rather than changing the tile, and leave the seams alone.
Finally, keep the mockup honest about what it is. A flat screen preview has no sheen, no relief, no weave, no thickness. A satin-finish wall surface under a hard key light can flare where your mockup showed calm colour. A printed textile brings its own much finer weave, which can interfere with your repeat in ways nobody predicts from a JPEG. A very fine repeat viewed from a distance can shimmer. None of that is a reason to skip the mockup; it is a reason to label it correctly. The preview settles composition, scale and rhythm. Whether the material behaves under a camera is a question for the people who own the material, and it is a separate question with a separate answer.
What you should be holding at the end
An editable tile, with its continuation copies placed as real objects and its tile dimensions written down. A multi-tile sample large enough to show the field. And a stated scale — this tile equals this many centimetres on this surface — so that the sample means something to anyone who receives it.
Then keep two verdicts apart. Is it seamless? is a geometric question with a right answer, and you can answer it yourself. Is it any good? is a compositional question about this café, this product and this frame, and it has no right answer, only a defensible one. A repaired seam does not vote on the second question, and a striking motif does not vote on the first.
That separation is most of the skill. Everything else — the coordinate arithmetic, the wrap copies, the field preview — exists to give you a version of the pattern you can actually look at, at a size that resembles the surface, so that the judgment you make is about the wall in the shot rather than the rectangle on your screen.
And when the tile is right and the field is right and the scale is right, the approval it earns is approval of a design proposal. It is not a manufacturing specification, not a licence to print, and not a promise about how a textured surface will behave under a light. Those are different conversations with different people, and the pattern is only the thing they will be talking about.
Frequently asked questions
Why can a technically seamless repeat still look wrong on a wall?
A repeat pattern is the relationship between one tile and its neighbours, not a picture of a motif. The join can be invisible and the wall still show a row of dots running straight across it, a cluster crowding one corner, or a rhythm that taps the eye. Those are field problems: accidental stripes, conspicuous clusters and empty patches appear when many tiles are seen together, not when one tile is judged alone.
How do you handle a motif that crosses the tile edge?
Place continuation copies as real objects. If a leaf overhangs the right edge by six units, place a copy translated left by exactly one tile width so the missing sliver enters from the next tile's left edge. If a motif overlaps a corner point, you need three copies: shifted left by one tile width, down by one tile height, and both. Set the tile size deliberately with a rectangle carrying the intended dimensions; otherwise a stray point can silently change the repeat. Illustrator can define and apply a pattern from selected artwork, but that step says nothing about whether the join is good.
How do you fix an accidental stripe made by aligned dots?
Change a relationship rather than sliding the whole family. If two dots share a y-coordinate, tiling produces a dotted rule across the wall. Moving both dots down by the same amount keeps the rule, just at a new height, because every copy slides with them. Give the dots different heights instead, so the line breaks into a zigzag. Then check that the moved dot has not opened a new boundary crossing. A tile-dimension change is blunt: it changes how far apart every repeated row sits, does not fix a single-family alignment, and invalidates boundary copies placed one tile width or height away.
What should you compare when testing the repeat at the surface's working scale?
Put the checked pattern on the surface at two plausible sizes and look at both. In the invented café wall, one tile = 20 centimetres gives 16 tiles across and a 2.4 by 4 centimetre leaf, so the wall reads as texture, but dot lanes every 20 centimetres may read as a woven band or camera wobble. One tile = 60 centimetres gives about 5.3 tiles across and a 7.2 by 12 centimetre leaf, so it reads as a motif and the lanes stop being a rhythm problem, though leaves that tall may fight a small product for attention. If the crop's phase matters, shift the pattern within the wall rather than changing the tile, and leave the seams alone. Keep the mockup honest: a flat preview has no sheen, relief, weave or thickness.
What should you hand over, and which two verdicts should stay separate?
Hand over an editable tile with continuation copies placed as real objects and its tile dimensions written down, a multi-tile sample large enough to show the field, and a stated scale such as this tile equals this many centimetres on this surface. Keep separate whether it is seamless, a geometric question you can answer yourself, from whether it is any good, a compositional question about this café, this product and this frame that has no right answer, only a defensible one. A repaired seam does not vote on the second question, and a striking motif does not vote on the first. Approval is approval of a design proposal, not a manufacturing specification, a licence to print, or a promise about how a textured surface will behave under a light.