People ask us for larger sheets. It is the most common request we get, and the answer is usually no — not because we do not want the sale, but because the picture is not there.
This is the arithmetic behind that answer. We publish it because a claim you cannot check is worth nothing.
What 300 dpi actually means
Dots per inch is not a property of a file. It is a relationship between a file and a size. The same 3,000-pixel image is 300 dpi printed 25 cm wide and 150 dpi printed 50 cm wide. Double the print and you halve the resolution. Nothing about the file has changed.
The fine art production houses that print for us specify artwork at “at least 300 DPI at the final print size”. Those last five words are the whole point. 300 is also roughly the ceiling of what matte fine art paper can resolve: ink spreads a little into the fibre, and detail finer than that does not survive the meeting. Above 300 you gain file size and nothing else. Below it you begin to lose the picture — gently at first, then visibly.
The trap
Any file can be made to say 300 dpi. Enlarge it in software, the pixel count rises, the preflight check turns green, and the file is now compliant by the number. But no information was added. The software invented the new pixels by averaging the old ones. What arrives on paper is a soft, faintly plastic version of the same image — convincing on a screen at 25 % zoom, disappointing at arm’s length on a wall.
The green badge measures the file. It cannot measure what is in the file. That is the single most useful thing to know about print resolution, and it is how a shop can sell a metre-wide poster made from a 2,000-pixel scan without technically saying anything untrue.
There are two further versions of the same trap, and both caught us.
A catalogue record can quote a different size from the one the server hands you. Sargent’s Olive Trees, Corfu is listed at 8,000 × 6,000 pixels in the Art Institute’s catalogue data; the file its image server actually delivers is 6,510 × 4,628.
And a server can cap what it will send in one piece. The National Gallery of Art in Washington describes its Monet Woman with a Parasol as 12,666 × 15,698 pixels, and that is true — but ask for the whole image and you get 3,305 × 4,096, because the delivery is capped at 4,096 pixels on the long side. Print from that and you are working at a quarter of the resolution you think you have. The way through is to ask for the picture in numbered regions and reassemble it — twenty tiles, in that case, joined on exact pixel boundaries. We measure the file we print from, never the number in the record.
Working backwards instead
We never scale a file up to reach a size. We count the pixels across the printed area of the original, divide by 300 and multiply by 2.54. That gives the widest honest image in centimetres. The sheet is then chosen around it, with the white border we use throughout.
Hiroshige’s Maples at Mama is the clearest case. The largest digitisation any institution publishes of that print — Minneapolis, accession 96.146.193 — carries 3,167 pixels across the printed block. Divided by 300, multiplied by 2.54: 26.8 cm. Our 40 × 50 sheet holds the block at 22.2 cm, its true size, with room to spare. A 50 × 70 sheet would need it 42 cm wide, and that print would have run at 156 dpi. So there is no 50 × 70.
What the rule cost us in a single week
In late August 2026 we re-measured every size in the catalogue against 300 dpi rather than the 150 we had worked to before. Nine of twenty-five were under the line. Three sizes left the shop and are not coming back.
Dürer’s Rhinoceros at 90 × 60 cm. Cleveland’s file gives 4,877 pixels across the block, and it is the largest that exists — Chicago’s sheet is trimmed, the National Gallery’s is smaller. At 90 × 60 that is 193 dpi. The work now stops at 50 × 40.
The Thiollier photograph at 50 × 50 cm. 2,880 pixels, and no better source traced anywhere. 174 dpi. It is now a single format, 30 × 30.
Maples at Mama at 50 × 70 cm. 156 dpi, as above.
These were not slow sellers we were relieved to drop. The large format is the profitable one.
It still happens. Monet’s late Irises — a canvas two metres square, painted at Giverny when he could barely see — comes to us at 5,407 pixels. That is 45.8 cm of honest print and no more, so the work is offered in one size, 50 × 50, and there is no large square. On a 70 × 70 sheet the picture would float in twelve centimetres of white on every side, which is not a border, it is an apology.
The other move: find a better sheet
Six sizes were rescued rather than withdrawn, by going back to the source and asking whether a better one existed.
All five Anna Atkins algae cyanotypes had been built from the Metropolitan Museum’s copy of Photographs of British Algae. But the New York Public Library holds the copy Atkins gave to Sir John Herschel, and has digitised it at roughly 1.7 times the Met’s resolution on every edge. The enlarged sheets went from about 170 dpi to between 324 and 336. Same work, better file, nothing conceded.
Then something we had not expected. Hokusai’s Great Wave had also come from the Met. The Art Institute of Chicago’s impression is digitised 2.6 times larger — and it is a different picture. The Met’s sheet has lost its pink dawn sky to fading. Chicago’s still has it.
That is worth dwelling on, because it is not a technical detail. A woodblock print is not one object with several photographs of it. Every impression is a separate pull from the blocks, printed in a different year, kept in a different drawer, exposed to different light. Some are early and strong, some late and tired, and nearly all have faded unevenly — the reds go first. Choosing which museum’s copy to print is a curatorial decision. File size is only where it starts.
We moved Sudden Shower over Shin-Ōhashi Bridge to Chicago for the same reason, and Maples at Mama to Minneapolis: that impression still carries the band of pink at the horizon which the Met’s has lost.
When the file, not the border, decides the sheet
The rule cuts the other way too. Carl Blechen’s Palm House on the Pfaueninsel is an oil painting 135 × 126 cm, and Chicago’s photograph of it is 6,050 pixels across. On our 70 × 70 sheet the standard border would put the picture at 58.8 cm wide, which needs 6,945 pixels. We do not have them. So the picture is set at exactly 300 dpi instead — 51.2 cm — and the sheet keeps a wider white margin at the sides than anywhere else in the catalogue. The border gave way, not the resolution. That is the trade the rule always makes.
The fern sheets show the same rule choosing a size rather than removing one. The fern cyanotypes Anna Atkins made with Anne Dixon in 1853 come from the Getty at between 3,800 and 4,000 pixels across the blue paper. At original size — 19.4 cm, the width the museum catalogues — that is around 500 dpi, more than the paper can use. The obvious second size would have been 50 × 70, to match the five Atkins algae sheets. It would have put the image 42 cm wide at about 245 dpi, and so it does not exist. The sheet is 40 × 50 instead: roughly 31 cm wide at 322 dpi, an enlargement of about 1.6× rather than 2.2×. The line between the two sits near 33 cm, and it is not a matter of taste.
Those sheets also carry a measurement we could have hidden and did not. The Getty catalogues each leaf as 25.4 × 19.4 cm, a ratio of 1.31. Measured on the digitisations, the coated blue paper runs to between 1.34 and 1.39 — it continues below the caption, where the catalogued image height stops. The catalogue is describing the image; we are printing the object. So we anchor the scale on the catalogued width, print the paper whole, and say on each page that the two numbers differ and why.
And sometimes the paper is the limit rather than the file. Van Gogh’s The Poet’s Garden comes to us at 12,637 pixels across — enough for a print 107 cm wide at 300 dpi. We stop at 63.8 cm, because that is what fits on an 80 × 60 sheet with a border. Washington’s photograph of Rousseau’s Equatorial Jungle is more extreme still, at 23,548 pixels: it would carry a print two metres wide, and we print it at 54. There is no virtue in a number you cannot hang.
Where the catalogue stands
Thirty-six works, sixty-nine sizes, every one of them at or above 300 dpi at its final print size. Nothing in this shop is upscaled.
The rule from here is the one that produced all of the above: the format follows the source. If a work is offered in only one size, that is why. If a size you had seen has gone, that is why too.
And if you find a better file than the one we used — a larger scan, a fresher impression, an institution we missed — write to info@museartis.com. We will look, and if you are right we will rebuild the product. Every product page names the collection and the accession number of the sheet we printed. That is not decoration. It is there so you can go and check.
0 comments