Why Your Prints Look Duller Than Your Screen: RGB, CMYK and Colour for Print-on-Demand

Every print-on-demand seller hits this once: the design looks electric in Photoshop, the sample arrives, and the colour is flat. Nothing is broken. You are looking at two physically different things — a screen that emits light and a shirt that reflects it — and expecting them to match.

Here is what actually happens to your file between upload and print, and the small number of decisions that change the outcome.

The one thing to get right

Send sRGB. Embed the profile. Do not convert to CMYK.

That is the whole answer for 95% of print-on-demand work. The rest of this article is why, and what to do about the colours that still won't behave.

What RGB and CMYK actually describe

RGB is additive: start with black, add red, green and blue light. All three at full strength gives white. Your monitor works this way, so do phone screens and the Etsy listing image a buyer sees.

CMYK is subtractive: start with white paper, subtract light using cyan, magenta, yellow and black ink. More ink means less light bounces back. A poster works this way, so does a mug print and DTG ink on cotton.

The two describe overlapping but different sets of colours. The set a device can produce is its gamut. The sRGB gamut is bigger than a typical CMYK press gamut in the bright saturated regions — greens, cyans, oranges — and slightly smaller in a few deep cyan-blues. Where sRGB reaches past what the ink can do, something has to give.

That something is the colour engine, and it will happen with or without your involvement. Your only real choice is whether you decide how it happens, or the supplier's automatic converter does.

Why you should not convert to CMYK yourself

Traditional print advice says convert to CMYK before sending. That advice comes from offset printing, where you knew the exact press, paper and ink set in advance.

Print-on-demand is not that. Your file might be printed on:

• A DTG printer laying water-based ink into cotton fibres
• A sublimation press bonding dye into polyester
• A DTF transfer heat-pressed onto a blend
• A large-format inkjet with six to twelve inks, not four, on poster stock

Each has a different profile. None of them is "generic CMYK". When you convert to a generic profile first, you compress your colour into a smaller box, and the supplier then expands and re-converts from that smaller box into their real one. Two lossy conversions instead of one.

Printful, Printify, Gelato and Gooten all publish the same guidance: send RGB. Their pipeline converts once, using the profile for the specific printer and substrate your order lands on.

The exception is a supplier who explicitly asks for CMYK with a named profile. Then use that exact profile — not "U.S. Web Coated (SWOP)" because it was the default.

Which RGB, though

Not all RGB is the same. Three come up:

The temptation is obvious: wider gamut means more colours, so surely a wider profile preserves more? It does — right up until something in the chain ignores the embedded profile and interprets the numbers as sRGB. Then every colour reads as under-saturated, and your vivid Adobe RGB file prints noticeably flatter than an sRGB version of the same artwork would have.

That failure mode is common enough, and the gain small enough, that sRGB is the correct default. Set it once in Photoshop under Edit → Color Settings → RGB: sRGB IEC61966-2.1, and always tick "Embed Color Profile" when exporting.

File format matters here too — PNG and JPEG both carry an embedded profile, and the choice between them comes down to transparency and compression rather than colour. That trade-off is covered in image file formats for Etsy and print.

The colours that never survive

Some colours are simply outside what ink on a given substrate can do. Print them and you get the nearest reachable neighbour, every time.

The repeat offenders:

• Neon and fluorescent anything. Neon green, safety orange, hot pink. Fluorescent inks exist but standard DTG and inkjet sets do not have them. Expect a noticeably calmer version.
• Pure saturated cyan. RGB #00FFFF has no ink equivalent. It lands somewhere around a bright but flatter sky blue.
• Bright orange and warm red at full saturation. These compress hard on DTG cotton in particular.
• Deep blacks on garments. A rich #000000 on screen becomes a dark charcoal on cotton, because ink soaks into fibres and scatters light.
• Very light pastels on white. Below roughly 5% coverage, DTG can drop out entirely and leave you with bare fabric.

Sublimation is the outlier in the other direction: because dye bonds into polyester rather than sitting on top, it holds saturated colour far better than DTG does — but only on high-polyester white or light substrates. The trade-offs there are set out in the sublimation printing guide.

Soft proofing: seeing the problem before you order a sample

Soft proofing simulates a print profile on your monitor. It is not perfect — you are still looking at emitted light — but it catches the big misses.

In Photoshop:

1. View → Proof Setup → Custom
2. Pick the supplier ICC profile if they publish one. If not, U.S. Web Coated (SWOP) v2 is a reasonable pessimistic stand-in for DTG.
3. Rendering intent: Relative Colorimetric with Black Point Compensation on. Perceptual if the design is a photo with lots of out-of-gamut area.
4. View → Proof Colors (Ctrl/Cmd+Y) to toggle between your file and the simulation.

Then turn on View → Gamut Warning (Ctrl/Cmd+Shift+Y). Out-of-gamut areas flood grey. Those are the regions that will shift.

The fix is almost never a redesign. Select the flagged colour, drop saturation 5–10%, nudge lightness a few points, and re-check. You keep the design and lose the disappointment.

Substrate changes everything

The same file prints differently depending on what it lands on. Two effects dominate:

Underbase on dark garments. DTG on a black shirt prints a white ink layer first, then colour on top. That base is never paper-white, so everything above it shifts darker and slightly warm. A design proofed on white and rolled out across a colour range will look consistently muddier on the dark options.

Fabric absorbency. 100% cotton takes ink differently from a 50/50 blend. Tri-blends are the least predictable — the polyester fraction rejects DTG ink, so colours read lighter and softer. That is a look some sellers want; it is only a problem when it is a surprise.

Practical rule: order one sample per substrate family, not per colourway. White cotton, black cotton, one blend. Three samples tell you almost everything about a fifteen-variant listing.

What this means for your listing images

There is a second colour problem that has nothing to do with printing: your mockup has to match the product.

If your mockup shows a vivid neon design and the buyer receives the muted printed version, that is a "not as described" complaint waiting to happen, regardless of who is technically right about gamut. The safe approach is to build listing images from artwork you have already gamut-corrected, so the mockup shows what ships.

This is one of the quieter arguments for generating mockups from the same corrected master file you send to the supplier, rather than from an earlier, brighter version. One source file, one colour reality. The master artwork sizing guide covers the file-structure side of keeping a single source of truth.

Resolution and colour are separate problems

Worth stating plainly, because they get conflated: colour shifts are a gamut problem, softness and pixelation are a resolution problem. A file can be perfectly sharp and badly out of gamut, or beautifully coloured at 72 DPI and unusable. Fixing one does nothing for the other. Resolution requirements per product type are in the print resolution and DPI guide.

The working process

When to stop worrying

Colour management rewards a couple of hours of setup and then stops paying. Once your working space is sRGB, your export embeds the profile, and you have physically seen one sample per substrate, you have removed nearly all of the surprise.

The remaining variance — batch-to-batch ink differences, one printer running slightly warm — is real, small, and outside your control. Chasing it costs more than it returns. Get the file right, see one sample, and put the time into listings that actually convert instead.