DPI, PPI, and What Actually Controls Print Size
Published 6/16/2026 · 10 min read · Image & design tools
Daniel Okonkwo — Front-end developer and tech writer at OneKitly
Web performance · File formats
Checked against 5 sources
An image is W×H pixels and nothing else. The DPI figure stored inside the file is metadata — a note telling a print dialog what default size to propose — and rewriting it alters no pixel data whatsoever. We proved it on a 311,380-byte JPEG: flipping the stored density from 72 to 300 changed exactly four bytes, left the file size identical to the byte, and the decoded 4,320,000-byte pixel payload hashed the same before and after. The only thing that changed was the number a print dialog reads. Printed size is pixels ÷ PPI. A 3000×2000 file at 300 PPI prints 10 × 6.67 in; at 150 PPI, 20 × 13.33 in; at 72 PPI, 41.67 × 27.78 in. Same file, same pixels, three answers. The rule of thumb of 300 for print comes from eyesight: one arcminute of visual acuity at a 10 in reading distance corresponds to about 344 pixels per inch, so 300 sits just under the limit of what you can resolve. At six feet — poster distance — the same arithmetic gives roughly 48 PPI, which is why 150 is comfortable there. Only resampling changes pixels; changing the DPI tag never does.
The DPI figure inside an image file is a label. It changes no pixels. Printed size is pixels divided by PPI — and DPI, properly used, means something else entirely: the printer's dots.
The DPI number in a file is a sticky note
Every common image format has somewhere to write a resolution. JPEG carries it in the JFIF header as an X and Y density plus a units flag; Exif repeats it as XResolution, YResolution and ResolutionUnit; PNG stores it in the optional pHYs chunk, in pixels per metre rather than per inch. None of these fields touches the image data. They sit beside it, in the header, and every one of them can be rewritten without decoding a single pixel.
We measured what a change costs. Taking a 1600×900 JPEG of 311,380 bytes and rewriting its stored density from 72 to 300 dots per inch modified exactly four bytes. The file size afterwards was 311,380 bytes — identical, to the byte. Decoding both versions to raw pixels gave two payloads of 4,320,000 bytes with the same SHA-256 hash. The only other difference anywhere in the decoded output was in four header bytes recording pixels per metre: 2,835 for 72 DPI, 11,811 for 300. Four bytes in, four bytes out, zero pixels touched.
So when a print shop or a stock library says "we need 300 DPI", they are almost never asking you to edit that field. They are asking for enough pixels that the image reaches 300 pixels per inch at the size they intend to print. Setting the tag to 300 on a small file satisfies the letter of the request and none of its substance — and any competent pre-press check will spot it immediately, because the pixel count is the thing they actually measure.
The only equation: printed size = pixels ÷ PPI
Take a 3000×2000 file. At 300 PPI it covers 3000 ÷ 300 = 10 in across and 2000 ÷ 300 = 6.67 in down. At 200 PPI: 15 × 10 in. At 150: 20 × 13.33 in. At 100: 30 × 20 in. At 72: 41.67 × 27.78 in — which is why a file tagged 72 lands in a print dialog proposing something absurdly larger than the page and gets scaled down to fit. Nothing in the file changed between those five answers. Only the divisor did.
Run it the other way when you are commissioning or shooting. A full-bleed US Letter page at 8.5 × 11 in and 300 PPI needs 8.5 × 300 = 2550 pixels across and 11 × 300 = 3300 down. A tabloid sheet at 11 × 17 in and 200 PPI needs 2200 × 3400. An 18 × 24 in poster at 150 PPI needs 2700 × 3600. Those are the numbers to put in a brief — not a DPI figure, which tells a photographer nothing about how much sensor they need to fill.
DPI is the printer's word; PPI is the image's
This is where most of the confusion is manufactured. PPI counts image pixels per inch of paper. DPI counts the ink or toner dots the printer actually lays down per inch, and a printer lays many dots per pixel, because a single dot of ink is one colour at full strength — it cannot be 42% cyan. To make 42% cyan the printer clusters dots and lets your eye average them. That is why inkjet specifications quote figures like 1200 or 2400 DPI while the image feeding them is 300 PPI: at 1200 DPI the printer has 1200 ÷ 300 = 4 dots per pixel along each axis, so 16 dots to render one image pixel. At 2400 DPI it has 64.
Offset printing counts differently again, in lines per inch — the frequency of the halftone screen. The trade rule of thumb is to supply about twice the screen frequency in image pixels, so a 150 lpi screen wants roughly 300 PPI and a fine 175 lpi screen wants around 350. That is the actual origin of the famous 300 figure, long before anyone had a phone with a camera in it. Once you separate the three units, the sentence "my image is 72 DPI" stops being a problem statement and becomes a category error: an image has no DPI, only pixels and a suggestion.
How much resolution the eye can actually use
Normal visual acuity resolves detail down to about one arcminute. Convert that into pixels at a given distance and you get the honest ceiling. At a 10 in reading distance, one arcminute subtends 10 × tan(1/60°) = 0.00291 in, so the eye tops out near 344 pixels per inch. The 300 PPI convention sits just below that, which is exactly why it looks sharp and why 600 PPI on a photo print is money spent on something nobody can see.
Push the distance out and the requirement collapses fast, because the ceiling is inversely proportional to distance. At six feet — the distance at which most people actually look at a poster — the same arithmetic gives 1 ÷ (72 × tan(1/60°)) ≈ 48 pixels per inch. The trade convention of 150 PPI for posters is therefore roughly three times more than acuity alone demands. That headroom is deliberate: somebody always walks up to a poster and inspects it from a foot away, and 150 survives that inspection while 48 would not.
Resampling is the operation that does change pixels
Resampling rebuilds the image at a different pixel count. Downsampling averages neighbouring pixels away and is generally safe. Upsampling has to invent pixels that were never recorded: going from 3000×2000 to 5000×3333 takes 6,000,000 real pixels to 16,665,000, meaning 10,665,000 of them — 64% of the output — are interpolated guesses. A good algorithm makes those guesses plausibly, but no algorithm recovers detail the sensor never captured. Resampling upward buys you a file that satisfies a pixel-count requirement; it does not buy sharpness.
The distinction matters most in the dialogue box where the two operations sit side by side. In every serious editor the image-size panel has a resample checkbox. Leave it off and typing a new DPI value only rewrites the metadata and updates the displayed print size — pixel count untouched. Switch it on and typing the same value rebuilds the image. The panel looks the same either way; the file that comes out does not.
What to do in practice
Decide the physical size first, multiply by the PPI the viewing distance justifies, and check whether your pixel count reaches it. If it does, set the DPI tag to that value so the print dialog opens at the right size and nobody downstream has to guess. If it does not, you have three honest options: print smaller, accept a lower PPI because of the viewing distance, or reshoot. Upsampling is a fourth option only when the shortfall is small.
For anything that will only ever be seen on a screen, ignore the field entirely. Browsers, presentation software and messaging apps all discard it. A 1600-pixel-wide image is 1600 pixels wide whether its header says 72, 300 or nothing at all, and the CSS or the layout decides how many of the viewer's own device pixels it occupies.
| Viewing distance | Typical output | PPI to aim for | Pixels a full page needs | Largest print from 3000×2000 |
|---|---|---|---|---|
| In the hand, about 10 in | Photo print, magazine page | 300 PPI | US Letter 8.5 × 11 in → 2550 × 3300 px | 10 × 6.67 in |
| On a desk, arm's length | Report cover, tabloid chart | 200 PPI | Tabloid 11 × 17 in → 2200 × 3400 px | 15 × 10 in |
| Across a room, about 6 ft | Poster | 150 PPI | 18 × 24 in → 2700 × 3600 px | 20 × 13.33 in |
| Twenty feet or more, in passing | Exhibition banner | 100 PPI | 36 × 48 in → 3600 × 4800 px | 30 × 20 in |
| A screen, any distance | Web page, slide, app | The DPI tag is ignored entirely | Whatever the layout asks for, in pixels | 3000 × 2000 px, full stop |
Frequently asked questions
- My printer asked for 300 DPI. Can I just set the field to 300?
- You can, and it will change nothing about the image. What they need is enough pixels for 300 per inch at the finished size. Multiply the intended width in inches by 300 and check your file has at least that many pixels across. If it does, setting the tag is a useful courtesy that makes the print dialog open correctly. If it does not, setting the tag simply hides the shortfall until someone checks.
- Why do people say web images are 72 DPI?
- It is a fossil. Early Macintosh screens were built so that one screen pixel matched one typographic point, and there are 72 points to an inch, so software of that era wrote 72 into the header by default. Modern displays are nowhere near 72 pixels per inch, and browsers have never read the field anyway. A web image tagged 72, 300 or 1 renders identically.
- Does changing DPI make the file bigger or smaller?
- Neither, if you are only editing the metadata. In our test the file was 311,380 bytes before and 311,380 bytes after, because the density is stored in fixed-width fields — the same four bytes hold 72 or 300. Size changes only when the pixel count changes, which means resampling, or when the file is re-encoded and the compressor produces a different result.
- Is 150 PPI really enough for a poster?
- For its intended viewing distance, comfortably. At six feet an average eye resolves around 48 pixels per inch, so 150 is roughly three times the theoretical requirement. The margin exists because posters get inspected up close by somebody standing right in front of them. If your poster carries fine type or a QR code that people will approach to read, treat those elements as reading-distance content and design to 300.
- My phone photo says 72 DPI. Is it low quality?
- The tag tells you nothing about quality. Look at the pixel dimensions instead. A 4032 × 3024 phone photo tagged 72 prints a 13.44 × 10.08 in image at 300 PPI — comfortably more than a full page. The same file tagged 300 would contain exactly the same pixels. Judge the file by its pixel count, its sharpness and its compression history, never by the density field.
Articles you may find interesting
All guides →Related tools
Sources
- W3C — Portable Network Graphics (PNG) Specification (Third Edition), pHYs — physical pixel dimensions
- ITU-T — Recommendation T.81 — the JPEG standard, and the JFIF density fields carried alongside it
- CIPA — DC-008 Exchangeable image file format for digital still cameras (Exif) — XResolution, YResolution, ResolutionUnit
- ISO — ISO 216 — Writing paper and certain classes of printed matter: trimmed sizes, A and B series
- MDN Web Docs — Image file type and format guide
Spotted a mistake in this article?