Daniel's been staring at a printout that looks nothing like his screen and he's got questions. A lot of them. Why screen colour never matches print, what ICC profiles are actually doing at conversion, why rich black isn't just one hundred percent K, why the same file looks different from two different presses, the trap of designing in RGB and only catching the shift at proofing, bleed and safe margins and why trimming is never precise, resolution for books versus banners, how paper stock and coating change everything, and how the failure modes differ across photographs, creative assets, roll-up banners, and books. He basically sent us the entire frustration of anyone who's ever hit print and gotten back something that looks like it was developed in a different decade.
And the thing is, every single one of those questions traces back to one fundamental problem. A screen emits light. Ink absorbs it. Those are two completely different physical processes, and pretending they're the same thing is where the whole house of cards starts wobbling.
So let's start with the fundamental lie your monitor tells you every single day.
Right. RGB is additive. Your screen starts from black — no light — and adds red, green, and blue photons until you get white. Maximum red plus maximum green plus maximum blue equals pure white. It's literally building colour by throwing more light at you.
And ink does the opposite.
Ink is subtractive. You start with white paper reflecting all the light back at your eye, and you lay down cyan, magenta, yellow, and black inks that absorb specific wavelengths. Cyan ink absorbs red light and reflects green and blue. Magenta absorbs green and reflects red and blue. Yellow absorbs blue and reflects red and green. Stack them all together and in theory you get black because all the light is absorbed. In practice you get a muddy brown, which is why we have the K — the black plate — to actually give you a real black.
So one system starts from nothing and adds, the other starts from everything and takes away. They're not just different processes — they're inverses.
And that inversion means their colour gamuts — the total range of colours each system can produce — don't overlap completely. RGB covers a much wider territory than CMYK, especially in the vivid blues, the electric oranges, the neon greens. Those colours that glow on a monitor? They have no ink equivalent. They simply don't exist in the printable world. The physics of pigment on paper cannot produce them.
Which means something has to give.
Something always gives. Every time you convert from RGB to CMYK, you're making a decision about what to sacrifice. And that decision is made by the colour profile.
So what is actually happening inside one of those ICC conversions? Because I think most people — myself included until I dug into this — treat profiles as a sort of magic black box. You pick one from a dropdown, the colours shift, and you hope for the best.
An ICC profile is essentially a translation dictionary. It describes how a specific device — your monitor, your printer, your press — reproduces colour. It maps out that device's gamut, its white point, its tone response curve. When you convert from one profile to another, the software uses a profile connection space as a middleman. Usually it's something like CIELAB, which is a device-independent colour space designed to represent colour the way humans perceive it.
So the RGB values get translated into this universal colour language, and then translated again into the nearest thing the printer can actually produce.
And here's where the compromise happens. The profile has to decide what to do with colours that fall outside the destination gamut. These are called out-of-gamut colours, and there are different strategies — called rendering intents — for handling them.
This is the part where the same file produces different results depending on which software you use, isn't it?
It is. Perceptual rendering intent squashes the entire source gamut down to fit the destination gamut. It preserves the relationships between colours — so a gradient still looks smooth — but every colour shifts a little, even ones that were in-gamut. Relative colorimetric clips out-of-gamut colours to the nearest reproducible colour and leaves everything else alone. It's more accurate for colours that fit, but it can produce banding in gradients where colours suddenly hit a wall.
So one preserves the feel, the other preserves the accuracy of what can be preserved.
And then there's saturation, which prioritizes vividness over accuracy — useful for pie charts and presentation graphics where you want punch, not fidelity. And absolute colorimetric, which tries to reproduce exact colour values including the paper white point, which almost nobody uses in practice because it makes your print look like it has a tinted background.
So when someone says "I converted to CMYK and my blues went muddy," what they're really saying is their profile chose a rendering intent that prioritized something other than what they wanted.
Or the blue was simply outside the CMYK gamut and no rendering intent could save it. A vivid electric blue — RGB values around zero, one hundred, two fifty-five — that's a colour your monitor can produce by cranking the blue subpixel to maximum. There is no combination of cyan and magenta ink that reflects only blue light at that intensity. The profile will either dull it to a navy or shift it toward purple, and that's not a failure of the profile — it's a failure of physics.
So the monitor is the liar here. Not the printer.
The monitor is absolutely the liar. It's showing you colours that can't exist on paper, and you've been looking at them for hours while you designed the thing. The printer is just the messenger.
That's a hard thing to accept when you're holding a print that looks like someone drained all the life out of it.
And it gets worse, because even within the printable gamut, different presses produce different results. Let's talk about why the same file looks different on two different machines.
Before we get there — rich black. Daniel asked about it specifically. Why isn't one hundred percent K enough?
One hundred percent K uses only the black ink plate. On coated paper it can look decent, but on uncoated stock it often comes out looking flat and slightly greyish, because the black ink alone doesn't fully absorb all the light. The paper texture scatters some of it back. Rich black adds percentages of cyan, magenta, and yellow underneath the black — a common recipe is forty percent cyan, forty percent magenta, forty percent yellow, and one hundred percent black. That extra ink underneath absorbs more light across the spectrum and gives you a deeper, denser black.
But you can't just max out all four plates.
No, and this is where people get into trouble. Most presses and paper stocks have a total ink coverage limit — typically around three hundred percent. If you set all four plates to one hundred percent, you're at four hundred percent total coverage, and the ink won't dry properly. It'll offset onto the next sheet, it'll smudge, it'll look glossy and wet even after it's supposedly dry. Registration also becomes a problem — if the paper shifts slightly between plates, you get coloured fringes around your black text.
So rich black is a recipe, and the recipe has limits.
And the recipe changes depending on the press and the stock. A rich black that works beautifully on a Heidelberg sheetfed press with coated stock might be a disaster on a web press running uncoated newsprint. Which brings us to why the same file differs across presses.
This is the part where the translation dictionary isn't enough.
An ICC profile is a model of a device at a specific moment in time. It's a snapshot. But a real printing press drifts. Ink density varies — the press operator adjusts the ink keys, and those adjustments change how much ink reaches the paper. Dot gain — the physical spreading of the ink dot as it hits the paper — varies with temperature, humidity, and press speed. The paper itself absorbs differently from batch to batch. The profile says "this is how this press behaves under these conditions," but the actual press on Tuesday afternoon is slightly different from the press on Monday morning.
So calibration is a snapshot, not a guarantee.
It's a best effort. A well-run print shop recalibrates regularly and runs colour bars on every sheet so the operator can check densities in real time. But there's always drift. The goal isn't perfect match — it's match within tolerance. For most commercial printing, a Delta-E of two to three is considered acceptable. Below one is essentially imperceptible. Above five, most people can see the difference.
And a brand manager looking at their corporate red will spot a Delta-E of two from across the room.
Every single time.
Now, the practical traps. You mentioned the RGB trap earlier — designing in RGB and converting at proofing.
This is the single most common mistake. A designer spends weeks working on a layout in RGB — Photoshop, Illustrator, whatever — and the colours look glorious. Vibrant blues, rich gradients, glowing oranges. Then they send it to the printer, the printer converts it to CMYK, and the proof comes back looking like someone threw a grey blanket over it. The designer is furious. The printer says "you sent an RGB file." And they're both right to be frustrated, because nobody caught it earlier.
The fix is working in CMYK from the start.
Or at minimum, soft-proofing. Soft-proofing uses the ICC profile of the target press and paper to simulate on your screen what the print will actually look like. It's not perfect — you're still looking at emitted light, not reflected light — but it's close enough to catch the big problems. You toggle it on and suddenly your electric blue sky is a muted teal, and you can adjust before you commit.
And hard proofs are the final check.
A hard proof — an actual physical print on the actual stock, or a certified contract proof from something like a Fuji or Epson proofing system — is the closest you'll get to the final result. But they're expensive and they take time. Most jobs don't get one unless the budget and schedule allow for it.
Bleed and safe margins. Daniel asked about trimming precision.
Trimming is never precise. Paper shifts as it moves through the cutter. Blades dull over the course of a run. A stack of paper compresses differently at the top than at the bottom. The industry standard is to expect about one to two millimetres of variation. So you extend your artwork three millimetres — or an eighth of an inch — beyond the trim line. That's the bleed. If you don't include bleed and the cutter drifts even half a millimetre, you get a white sliver along the edge of your print.
Safe margins are the inverse.
Keep all critical content — text, logos, anything you don't want cut off — at least three to five millimetres inside the trim line. Nobody wants their phone number bisected by a blade.
Resolution. This is one where I think the conventional wisdom — "always use three hundred DPI" — is actually wrong in a lot of cases.
It's wrong because resolution should match viewing distance. A book page held at arm's length — about thirty to forty centimetres — needs around three hundred dots per inch for the human eye to perceive it as continuous tone. That's the standard. But a roll-up exhibition banner that someone stands two feet from? You can get away with one hundred to one fifty DPI. The viewing distance compensates for the lower resolution.
The failure pattern is different.
A banner printed at seventy-two DPI looks pixelated if you walk right up to it, but from ten feet away — which is where most people will see it — it looks fine. A book page at one fifty DPI looks soft at reading distance. Your eye can resolve the difference because you're close enough to see the individual dots. Same resolution, completely different result depending on where the viewer is standing.
The person who insists on three hundred DPI for a billboard is just wasting file size and processing time.
A billboard viewed from fifty metres away can be printed at fifteen DPI and look perfectly sharp. The math is straightforward — it's the angular resolution of the human eye, about one arcminute. Beyond that, your eye can't distinguish two points as separate.
Paper stock. This is where the same file on different paper looks like a different print job entirely.
Coated paper — whether glossy or matte — has a smooth surface with a clay or polymer coating that sits on top of the paper fibres. The ink sits on that coating rather than soaking into the fibres. You get sharper dots, less dot gain, and more vibrant colour because the light reflects cleanly off the surface. Uncoated paper is absorbent — the ink soaks in and spreads, like writing on a paper towel with a felt-tip pen. The dots get larger, colours darken and muddy, and fine details lose definition.
The same rich black recipe produces a deep, crisp black on coated stock and a flat, slightly fuzzy dark grey on uncoated.
There's no profile in the world that can fully compensate for that difference, because the paper is part of the colour. The white point of the paper — how white it is, whether it's warm or cool — shifts every colour that sits on top of it. A cream-coloured uncoated stock makes everything look warmer and less saturated just by being the background.
Let's get into the formats Daniel asked about. Photographs, creative assets, roll-up banners, books. The tolerances and failure pattern are genuinely different for each.
Photographs are the hardest case for colour accuracy. The human eye is trained on natural colours — we know what a blue sky looks like, what skin tones should be, what grass looks like. If a landscape photo comes back with a muddy sky or orange skin tones, it's immediately wrong. The fix is to convert with a rendering intent that preserves the overall feel — usually perceptual — and accept that some individual colours will shift. The goal is a print that looks good on its own terms, not one that matches the screen exactly, because it can't.
Creative assets are the opposite problem.
Logos and branding are about absolute consistency. A corporate red that's off by a few Delta-E is a crisis. Coca-Cola red, Tiffany blue, Home Depot orange — these are legally protected colours in many cases. The fix is to use spot colours from the Pantone Matching System rather than process CMYK. A spot colour is a premixed ink — the press runs a fifth plate with the exact Pantone formula. It's not built from cyan, magenta, yellow, and black dots. It's the actual colour, mixed to specification.
Which is why brand guidelines always specify Pantone references.
Why a brand manager will reject a proof where the red shifted by five Delta-E even though most people wouldn't notice. The tolerance is zero because the brand is the colour.
Roll-up banners.
The failure pattern here is different entirely. Colour accuracy matters less because nobody is comparing the banner to a reference — they're seeing it in isolation at an exhibition hall under terrible fluorescent lighting. What matters is physical durability. The banner has to be rolled and unrolled dozens of times without the ink cracking or flaking off. The substrate matters — vinyl, fabric, whatever — and the ink has to be flexible enough to survive the abuse. Resolution can be lower because of viewing distance. Ink coverage can be higher because the material can handle it. But if the banner cracks along the fold lines after three uses, none of the colour accuracy matters.
And books.
Books are about consistency across pages and across copies. A three-hundred-page book printed in one press run needs uniform colour from page one to page three hundred. The press operator is watching the colour bars and adjusting throughout the run. But a reprint six months later — different paper batch, different humidity, different press conditions — might drift noticeably from the original. Readers notice a bad image. They don't notice a slight shift in the black density of body text from one copy to the next. The tolerance is moderate — good enough across the run, and close enough on the reprint that nobody complains.
The failure pattern for a photograph is colour accuracy, for a logo it's brand consistency, for a banner it's physical durability, and for a book it's uniformity.
All of them are symptoms of the same underlying problem: print is a physical process with inherent variability. The ink, the paper, the press, the humidity, the operator — every variable introduces drift. The question isn't "how do I eliminate the drift" because you can't. The question is "how do I keep the drift within a tolerance that's acceptable for this specific thing I'm printing."
That's why printers ask for PDFs with embedded profiles. They want the manifest, not just the cargo.
A PDF with embedded ICC profiles tells the printer exactly what colour space the file was designed in and what the intended output conditions are. Without that, the printer is guessing. They'll apply a default profile, and the result will be whatever their default happens to produce. Embedding the profile is saying "here's the dictionary I used — use the same one."
Which connects to what we talked about last time with PDF/X and why it mandates embedded output intents. The file format and the colour management are the same problem seen from two angles.
Right. The PDF carries the instructions, the profile carries the colour translation, and the press carries the physical reality. All three have to agree, or the print is wrong.
I want to go back to something you said earlier about the monitor being the liar. Because I think that's the single biggest misconception people have about this entire process.
It really is. People assume the screen is showing them the truth and the printer is getting it wrong. But the screen is the outlier. It's producing colours by blasting photons directly into your eyes. No reflective surface can match that intensity or that gamut. The screen is a light source. The print is a surface illuminated by ambient light. They're not the same category of thing.
The better the screen gets, the worse the mismatch becomes.
This is the cruel irony. Modern displays — HDR, wide gamut, OLED — can produce colours that are further from the printable gamut than ever before. A new MacBook Pro display covers the P3 colour space, which is significantly wider than sRGB. The colours look incredible on screen. They look even more impossible in print.
As display technology improves, the gap widens.
It does. Digital presses are narrowing the gap on the print side — things like HP Indigo presses with additional ink stations can hit a wider gamut than traditional offset — but the physics of ink on paper has a ceiling. You can't make pigment reflect light that it absorbs. The mismatch is fundamental.
Meanwhile, the person at home with a hundred-dollar monitor that's never been calibrated is designing a wedding album and wondering why the prints look like they were taken underwater.
Which is a perfect description of —
Hold on.
Hilbert: Nineteen ninety-three. I ran a small print shop. Hilbert's Quick Print. Two-colour Ryobi press, a Macintosh Quadra, and a complete ignorance of what I was doing. A customer brought in a wedding photo — his own wedding, shot on film, scanned on a flatbed scanner that cost about eighty dollars — and wanted a dozen prints for the family. The file was RGB. I didn't know what an ICC profile was. I sent it straight to the RIP and the prints came back looking like the happy couple had been married at the bottom of a swimming pool.
What did you tell him?
Hilbert: I told him the paper was defective. He believed me. I reprinted on a different stock — same result, obviously — and he paid for both runs. I still have one of the prints framed on the wall above my desk. It's a reminder that the customer doesn't know what's wrong, but they know something is wrong.
The monitor was the problem, not the press.
Hilbert: I learned that about two years later. Another customer — restaurant owner, wanted a specific shade of sunset orange for his menu. We ran three proofs, each one slightly different, and he rejected all of them. I was pulling my hair out. Finally I went to his office to look at what he was seeing, and he had this cheap TN panel monitor — you know the kind, where the colour shifts if you tilt your head half an inch — and it wasn't just uncalibrated, it was actively wrong. The white point was somewhere around nine thousand Kelvin. Everything looked blue. He'd been adjusting his design to compensate, pushing everything toward orange, and on a normal screen it looked like a traffic cone.
Did you tell him?
Hilbert: No. He was already convinced I was incompetent. Telling him his monitor was lying to him would have just made him angrier. I calibrated my own screen, eyeballed what I thought he actually wanted, ran a fourth proof, and he said "finally." Paid the invoice. Never used him as a reference.
You solved the problem by guessing what his screen was doing to the colours.
Hilbert: I solved the problem by understanding that he wasn't looking at the same file I was looking at. The RGB values were identical. The display was not. That's the thing about colour management — it only works if the whole chain is managed. One uncalibrated screen at the start and everything downstream is wrong, and the person at the start doesn't know it.
You still have the wedding photo.
Hilbert: Framed. It's a good reminder. Also a terrible photo. His wife is wearing a dress that on screen was ivory and in the print was the colour of weak tea. They're still married, as far as I know. I think about that sometimes.
The gap between what people see and what they get — it's been the same problem for thirty years, and the physics hasn't changed.
Hilbert: The physics hasn't changed. The screens got better, which made it worse. Now people are designing on phones with OLED displays that can show colours a printing press won't touch for another decade, and they think the print is broken. It's not broken. It's just paper.
The screen is the lie we've learned to live with.
Hilbert: That's it.
The next time a print job comes back wrong, the question isn't what went wrong. It's which compromise did the profile choose, and was the screen ever showing you something real in the first place.
For anyone preparing files — work in CMYK from the start, soft-proof with the target profile, embed your profiles in the PDF, include bleed, respect the safe margins, match resolution to viewing distance, and know your paper stock before you design a single pixel.
Or just accept that the print will look different and make your peace with it. That's also a valid strategy for a lot of jobs.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop, who apparently has a museum of printing disasters above his desk.
If you've got a weird prompt, send it to show at my weird prompts dot com. We'll be back soon.
