Custom QR Stamp Generator
A custom QR stamp has to survive being read by a machine, so this one tells you when your design has stopped being readable. Set it below and watch the verdict change as you go.
Build a custom QR stamp
Version 4, 33 modules across, plus the four-module margin on every side.
Arced text around a ring, code in the middle
Best for: Anything with a label over the middle, or stamped by hand
Costs you: Densest symbol, so each module is smaller at a fixed stamp size
This ink measures 14.2:1 against white paper. A reader wants dark against light, and anything under 7:1 is the first thing to fail on a coloured surface.
41 modules across, so each one lands at 0.93mm. Comfortably over the 0.5mm floor this page uses for a rubber die.
ExportBoth are free and unwatermarked. The PNG comes out 1000 pixels wide on white, because a code on a transparent background stops reading the moment it lands on something dark. The SVG is vector and transparent, which is the one to send a die cutter.
Everything is drawn in this tab. No link, contact card or file is sent anywhere.
Six designs to start from
Each one is a custom QR stamp that loads into the builder above. They differ in the thing that matters, which is how much of the code they spend, and on what.
How much of a custom QR stamp you can cover before it breaks
Every QR code carries spare copies of its own contents. Denso Wave, who invented the symbology, publishes four levels of it: L rebuilds 7 percent of the code, M 15 percent, Q 25 percent and H 30 percent. That is the number every article about putting a logo in a QR code quotes, and it is the number that gets people into trouble, because it counts codewords rather than surface area, and because of what a plate over the middle actually does to a symbol. A QR code interleaves its codewords across the whole matrix, on purpose, so that one blot cannot destroy any single block outright. The side effect is that a plate puts a little damage into a great many blocks at once, and correction is applied per block. A little everywhere is harder to survive than a lot in one place.
So we measured it instead. Every custom QR stamp in the table below was rendered to a bitmap and handed to a decoder, with the plate grown one step at a time until the decoder gave up. The payload is this page's own URL, the frame is the round one, and the render is perfectly clean, with no ink anywhere in it.
| Level | Published capacity | Widest plate that still read | Narrowest plate that failed |
|---|---|---|---|
L | 7% of codewords | 3% of the code covered | 4.3% of the code covered |
M | 15% of codewords | 9.2% of the code covered | 11.1% of the code covered |
Q | 25% of codewords | 12.3% of the code covered | Nothing this tool can draw |
H | 30% of codewords | 11.7% of the code covered | Nothing this tool can draw |
Read the L row twice. It gave out at 4.3 percent coverage against a published capacity of 7 percent, so the headline number was worth about three fifths of itself once the damage was concentrated in one square. M behaved the same way, failing at 11.1 percent against 15. Q and H were not broken by anything this builder is willing to draw, which is why the slider stops where it does.
Treat those as a ceiling a custom QR stamp will never reach in the real world, not a target. One payload, one decoder, one clean render, and no ink in any of it. The budget this page holds you to is 60 percent of the published capacity for exactly that reason: the smudges, the half-pressed corner and the phone held at an angle all come out of the same account, and they have not been charged to it yet.
The other half of the answer is where the plate sits rather than how big it is. The three heavy squares in the corners, their white separators, and the two dotted lines running between them are function patterns. A reader uses them to find the code and work out its grid before it decodes anything, so they are not covered by any correction level. Some readers cope with a nick out of one, so this is not a guarantee of failure, but moving from L to H buys you exactly nothing here, which is the part worth knowing. The plate on this page is capped so it cannot reach them.
There is a fourth kind, and it is the one that catches people out. Alignment patterns show up from version 2, and their job is to let a reader unwarp a picture taken at an angle. For small symbols there is one, and it sits down in a corner where nothing disturbs it. From version 7 they start landing near the middle instead, which is exactly where a centre plate goes. A long payload such as a full contact card needs a symbol that size, so on those designs a centre plate lands on one no matter how small you make it. That is not automatically fatal. A decoder looking at a flat scan usually copes; a phone held at an angle over a stamped page, which is the normal way one of these gets read, copes less well. The builder says so when it happens rather than quietly letting it through, and the honest fix is a shorter payload rather than a smaller plate.
The margin nobody draws
Denso Wave states the requirement in one sentence: a QR code requires a four-module wide margin at all sides of a symbol. It is not padding and it is not taste. A reader uses the empty band to tell where the code stops, and a design that crowds a ring or a line of text up against the last row of modules has taken part of the code away without touching a single module.
This is the most common way a custom QR stamp fails, because a stamp frame is a ring or a box drawn around something, and the natural instinct is to close the gap and make the artwork look tight. So the 4 module margin is not a control on this page. It is built into the layout, the frame is placed outside it, and the preview shows you the real proportion rather than a flattering one.
Ink, paper, and the pretty red one
The text stamps on this site offer a warm oxblood because it looks like a stamp should look. On a code it is the riskiest thing on the panel. A reader works on light against dark, and that red measures 5.9:1 against white paper where the near black measures 18.9:1. Both scan off a clean white sheet in good light. Only one of them still scans off a kraft mailer, under a warm bulb, through a phone camera that has decided to expose for the room.
Each ink swatch reports its own ratio, and a custom QR stamp under 7:1 gets flagged even when the geometry is perfect. The ratio is borrowed from the accessibility standard rather than the QR one, which asks for dark on light without naming a number, so use it to compare inks against each other rather than as a pass mark. The comparison is the useful part.
One more decision the tool makes for you. The exported PNG has white paper under it, while the SVG stays transparent. A transparent code dropped onto a dark background inverts, and an inverted code fails silently, so the raster export you are most likely to paste into a document carries its own background. The vector one does not, because that file is going to a die cutter who wants the shape.
How small a stamped code can get
The symbology has no minimum physical size, which is technically true and practically useless. What has a minimum is your die. Below roughly 0.5mm a module, ink creeping sideways out of the relief starts closing the light squares, and a closed light square is a wrong module rather than a dirty one. Correction can argue with dirt. It cannot argue with the code saying something else.
The arithmetic is short. Modules across, including both margins, times your floor. A short asset id at level H needs 37 across, so around 19mm of stamp. A full contact card at the same level runs past 60 across, so past 30mm, which is a large stamp to press by hand. That is the real reason to think about what you encode: a shorter payload is not tidier, it is physically smaller, and the builder above recalculates it while you type.
Two ways out when a custom QR stamp comes back too big to press by hand. Encode a short link on your own domain rather than a tracking URL with forty characters of campaign parameters, or drop from H to Q and accept a smaller margin for error. The tool will tell you what each one buys.
Sources: Denso Wave on error correction levels, Denso Wave on determining the code area and margin, ISO/IEC 18004, the QR Code specification. All three were read on 28 July 2026. The coverage table is our own measurement, produced by decoding this tool's output, and the method is described above it.
What people stamp these on
Four jobs come up over and over, and each one wants something different from a custom QR stamp.
Inventory and asset tagging
The easiest case, because an asset id is short, so the code is sparse and the stamp can be small. The rectangular frame suits it: code on the left, department and owner beside it in plain text, so a person can read the tag without a phone. Level H, dark ink, and stamp onto a white label rather than the equipment.
Event check-in
Read once, indoors, off a clean badge, by a phone held six inches away. This is the one case where Q is genuinely enough and the extra density of H buys you nothing. Keep the code and the event name on the same face so a stack of unstamped badges cannot be confused with a stamped one.
Business card seals
A vCard encodes a whole contact record, which is a lot of characters and a dense code. Either accept a wide stamp, or encode a link to your own page instead and let the phone fetch the details. The second option makes a far smaller code and survives you changing phone number, which the first one does not.
Document authentication
A code that points at a record on your own site, stamped on the paper copy. This is the case with the longest life: the page will be filed, photographed, photocopied and read back in two years, and every one of those steps spends part of the correction budget. Level H, no overlay, the darkest ink, and a URL short enough to keep the module size comfortable.
What this does not do
A custom QR stamp from this page is artwork. It is not a physical stamp, we do not sell one, and we do not know what your manufacturer's frames measure. It does not track scans, and the code it draws is static, so the destination is baked into the pattern and cannot be edited afterwards. If you need to change where it points, put a redirect you control in the middle.
It also cannot promise your custom QR stamp will scan. Nothing that draws a picture can. What it can do is refuse to draw the designs that certainly will not, tell you which budget you are spending, and be honest that the last stretch of that budget belongs to the ink and the hand rather than to the code. Press a test sheet on the stock you will actually use, scan it with the oldest phone in the office, and only then approve the die.
Two smaller limits. The stamp is drawn in Georgia rather than the display face used elsewhere on this site, because an exported file that references a webfont it cannot carry arrives looking nothing like the preview, and wherever the SVG is opened without Georgia the viewer will substitute something else. And the correction levels are the four the standard defines, so if you were hoping for something between Q and H, there is nothing there to offer you.
Questions people actually ask
Will a QR code stamp actually scan when it is stamped in ink?
Usually, on light uncoated paper, pressed flat. The failures are physical rather than digital. Ink spreads sideways out of a rubber relief, so the light squares between the dark ones narrow, and once one closes the reader sees a wrong module rather than a smudged one. Rocking the stamp instead of pressing it straight down smears one edge, which is the same problem. Oil-based ink on matte stock holds a fine relief better than a water-based pad on anything glossy. Design at level H, keep the printed module over half a millimetre, and stamp a test sheet before you stamp two hundred mailers.
How small can a custom QR stamp be and still work?
Work backwards from one module rather than from the stamp. Count the modules your data needs, add the four-module margin on each side, and multiply by the smallest square your die and ink can hold. This page uses half a millimetre as that floor, which is a working figure rather than a standard, because a rubber die inked by hand is not a laser printer. A short asset id at level H comes to 37 modules across with the margin, so about 19mm. A full contact card at the same level runs past 60 modules, so past 30mm, and the tool tells you which side of the line your design is on as you type.
Can I put my logo or text in the middle of the QR code?
Yes, and the middle is the only place it is safe to put one. The three big squares in the corners and the two dotted lines that link them are how a reader finds the code and works out its grid. They are not data, so error correction cannot rebuild them, and moving up a correction level protects them not at all. Some readers survive a nick out of one, but it is not damage you would choose to design in. The middle is data, which correction can rebuild, up to a point. The slider on this page caps the plate before it can reach a corner square and shows you what fraction of the code you are spending.
What error correction level should I use for a stamp?
H for a custom QR stamp, unless the code is so dense at H that the modules fall under your printing floor, in which case Q. The levels rebuild 7, 15, 25 and 30 percent of the code at L, M, Q and H, and the cost of a higher level is a bigger symbol for the same link, which means smaller modules at a fixed stamp width. Screens can afford L. A rubber stamp cannot: the ink, the paper and the hand all take their cut of the same budget before anyone points a phone at it.
Can I change where the code points later without ordering a new stamp?
Not with what this tool makes. The code here is static, meaning the link is encoded in the pattern itself, so a new destination is a new pattern and a new die. If you need to redirect it later, encode a URL you control that redirects, such as a short link on your own domain, and change the redirect rather than the stamp. That is the same thing commercial dynamic QR services sell, done on infrastructure you already own, and it costs you nothing per scan.
Does it work on kraft boxes and brown packaging?
Less well than on white, and the reason is contrast rather than colour. A reader wants dark against light. Brown kraft is already halfway to dark, so ink on it has less room to be darker still. Two things help: give the custom QR stamp the darkest ink on the panel rather than the one that matches your brand, and stamp onto a white label rather than the board itself if the scan matters. The tool shows the contrast ratio of each ink against white paper so you can see how much headroom you are giving away before you commit to a die.
Can I turn this into a real rubber stamp?
This makes the artwork for a custom QR stamp, not the stamp itself. Send a manufacturer the SVG rather than the PNG: it is vector, it is one flat colour, and it carries no fixed physical size, which is what a die cutter wants. Whether they accept SVG, what frame sizes they cut, and whether they can hold your module size in rubber are questions for them. Ask for a proof impression on your own stock and scan it before you approve the run.
Where to go next
Round stamp maker
Circular stamps with a double ring and curved text.
Custom stamp generator
Design any stamp from scratch: text, shape, ink and export.
Company seal maker
Corporate seals with an arced company name and founding year.
Date stamp generator
A dated band and a status word, in the format your reader expects.
Stamp templates
Editable layouts for office, wedding, teacher and address stamps.
Business logo stamp
Check whether your logo survives being inked, then build around it.
A note on the correction levels quoted throughout: L rebuilds 7%, M rebuilds 15%, Q rebuilds 25%, H rebuilds 30%. Those are Denso Wave's published figures, and the same constants drive both this sentence and the tool above, so the page and the picture cannot disagree.