Common file issues and how we fix them

The edge of any smooth-edged image is a ramp of 1 to 3 partly transparent pixels. Your screen blends them into the page. The press prints white ink under them, so on a dark shirt they show as a faint pale outline around the whole design.

Anti-aliasing is how software makes a curve look smooth on a grid of square pixels: the pixels along the edge are drawn partly transparent so they blend between the design and whatever sits behind it. On screen that background is your page, and the blend reads as a clean edge.

DTF has no “partly”. The press lays a white underbase under almost any pixel that is not fully transparent, then prints color on top. A 30%-opaque edge pixel prints as solid white ink with a thin coat of color: a pale rim, a few pixels wide, around everything. On a white shirt you cannot see it. On black, navy or red it is the outline everyone notices.

The press already pulls the white layer in by 0.004 inch to hide the first pixel of that ramp. At 250 ppi at print size a pixel is exactly 0.004 inch, so a normal one-pixel edge hides inside it. At lower resolution, or with a wider ramp, the white shows.

Defringe recolors the edge pixels to the design's own color. Edge cleanup turns the soft ramp into a hard edge with a single pixel of smoothing. The result is an edge the press can print exactly, with the white underbase tucked inside the color.

A halo has two parts. The edge pixels are the wrong color, a blend toward white or toward the old background, and they are partly transparent. Defringe fixes the first: each edge pixel takes the color of the solid pixels just inside it, so whatever prints at the rim is the design's color, not a pale mix. Edge cleanup fixes the second: pixels that are mostly transparent are dropped, pixels that are mostly opaque are made solid, and one pixel of smoothing is redrawn so curves stay smooth. That one pixel is what the press's choke covers.

Choke is the press-side version of the same idea: the white layer is drawn 0.004 inch inside the color layer, so white never peeks out and the print head can drift a little without showing a white line. We set that on the press. There is nothing for you to do about it in your file. A clean, hard-edged PNG at 250 ppi or more at print size makes all of this a formality.

A stair-stepped edge has no soft pixels at all, and the press prints every step exactly. The fix is to redraw the edge at print resolution, not to blur it where it is.

Jagged edges come from a magic-wand cut-out, a transparency hard-thresholded at low resolution, an auto-trace of a small image, or a small image enlarged by repeating pixels. Each step is one pixel of the original. At 150 ppi at print size that is 0.0067 inch, plainly visible on a curve or a diagonal. At 300 ppi it is 0.0033 inch and disappears at reading distance.

Blurring the edge as it stands makes it worse. Softening a 150 ppi stair turns each step into a three- or four-pixel ramp, which is the halo from the section above, now around the whole design. The stairs at least had no white showing.

What we do is re-snap the edge at print resolution: resample to the pixel grid the file will print at, redraw the edge with one pixel of smoothing on that grid, and harden the rest. The step becomes one print pixel, inside the choke. If the source is a screenshot of a logo, a vector redraw is the real fix, and that changes the shapes, so it is your call.

A speck is a pixel or two that survived a cut-out. It still prints as a dot of ink. On a dark shirt it shows, and at that size it may not bond. We remove isolated specks outside the design; marks inside the design are left alone unless you ask.

Specks come from a magic-wand or color-key cut-out that left islands the tolerance missed, from JPEG compression noise around edges, from screenshots whose “empty” areas carry leftover shadow or text smoothing, and from dust in scans. A two-pixel speck at 300 ppi is well under 1/100 of an inch across. That is still enough to fire white ink, so it prints. A mark that small carries almost no adhesive through a hot peel; it may bond, lift, or come off in the first wash. Either way it is a fleck on the garment that was not in the design.

Specks away from the design also make the transfer bigger: the print area includes every printed pixel, and you pay for that area.

As a rule, do not count on an isolated mark smaller than about 1/32 of an inch to stay on the shirt. Marks inside the design, such as distress texture, sparkle or halftone dots, are usually the design, so we leave them unless you tell us otherwise.

A JPEG has no transparency, so the white behind your logo is pixels, and pixels print, as a solid panel of white ink. Removing it changes how the design looks, so it is your decision. Our art team sees it during the check and asks.

JPEG cannot store transparency at all, and a screenshot or a flattened PNG has already filled the background in. To the press, the white rectangle is a design element: white underbase, white ink and adhesive, edge to edge, with the stiff hand of a solid fill. On a white shirt it shows as a glossy rectangle. On anything else it is a white box.

Color Knockout removes one flat color, within a tolerance, and leaves transparency behind. It is the right tool when the background really is one color and distinct from the art. It is the wrong tool when the same white is inside the design, in white text, highlights or a white keyline, because those go with it. That is why nobody should knock a background out unasked: a white circle logo that fills the frame and a logo on a white card look the same to software.

You can remove it yourself in the studio, with Color Knockout for a flat color or Background Removal for a photo or scene. Either way, expect a light rim where the art met the white. That is the fringe from section 02, and the cleanup handles it. Or ask us to do it: Help me with my art in the studio or the cart — how a request works.

A line that is too thin carries too little ink and adhesive to bond during a short press. A gap that is too narrow fills in. We look at stroke width at your print size during the check and flag it. We do not thicken lines without asking, because that changes the design.

Adhesive powder sticks only to wet ink. A hairline carries a thin bead of it, so it under-bonds and lifts on the hot peel or washes out. On dark shirts there is a second effect: the white underbase is pulled in by 0.004 inch on each side, so a thin positive line loses its white core and prints as translucent color over the shirt. On black it vanishes.

Gaps fail the other way. Ink spreads a little under heat, powder bridges narrow gaps, and press pressure squeezes molten adhesive sideways. The holes in small letters close up, and fine hatching turns to mud.

As a rule, keep strokes at 1.5 pt or wider and gaps between solid areas at 2 pt or wider, measured at the size you order. Text is the usual offender. We measure the stroke, not the font size, because a light face and a bold face at the same size are different lines. If a stroke is under about 1 pt at print size, we say so. Thickening it is a change to your design, so you decide.

What matters is pixels divided by inches at the size you order, and nothing else. 300 ppi at print size is the target. Below 175 the file will not hold at that size, and we talk before printing.

Take the width of your file in pixels and divide by the width you are ordering in inches. 3,000 pixels at 10 inches is 300 ppi. The same file at 20 inches is 150. Nothing in the file changes. Only the size you pick does.

What you see: at 300 ppi a pixel is 0.0033 inch, below what the eye separates even up close, and edges look drawn. At 250 they still do, and one pixel of edge smoothing sits inside the press's choke. At 200 a pixel is 0.0050 inch: fine at arm's length, slightly soft up close, small text a little heavy. At 150 curves step and small text loses its shape. The print head reproduces each pixel exactly, so blur and stairs print crisper than they look on your monitor.

The “DPI” in a file's metadata is a label, not a fact. Any app can write 300 into it without adding a pixel. We ignore it and count pixels against the size you pick, live, as you change the size. Below 175 ppi at size we stop and talk: an up-res (never more than 4×), a smaller size, or a better original.

Design in RGB. We convert to the press's inks with our own press profile at print time. Converting to CMYK yourself with a generic profile throws color away. A few colors cannot be reached in ink, and that is a color difference, not a defect.

Every design tool works in RGB, and the RGB you see on screen is the input our color tools and press profile expect. Converting to CMYK in your own software uses a generic offset-press profile that is narrower than our ink set, so you lose saturation the press could have printed, and it cannot be recovered. Send the RGB file. If your file is already CMYK, from a print shop's PDF for example, we keep its appearance. It gains nothing by being CMYK.

Some colors sit outside what pigment ink on a white underbase can make: pure RGB blue, which prints toward purple; fluorescent and neon colors; some very saturated oranges and greens; and certain brand colors. Our ink set narrows the gap but does not close it. Where a color cannot be reached, the print shows the nearest color ink can make. The transfer is fine. The color is not the one on your screen.

The color-risk check in the studio shows which areas are at risk before you order, so the conversation happens up front instead of after the transfers ship.

A soft shadow, a glow or a fade to transparent asks the press for 25% ink. Film has no 25%. It prints as solid white underbase with a thin coat of color: a pale, solid-looking patch instead of a fade. Halftoning turns the fade into dots the press can print.

On screen a 25%-opaque black shadow shows 75% of whatever is behind it. On DTF the white underbase fires under every pixel that is not fully transparent, and the color prints at 25% on top of it. The shirt never shows through, because there is ink under every pixel. On a black shirt the shadow becomes a grey fog around the design and a glow becomes a milky patch. On a white shirt these are nearly invisible, which is why the same file can look right on one order and wrong on the next.

The faint tail of a fade has a second problem: at very low opacity the color drops out before the white does, so the outer edge prints as white only, a thin, patchy film that is the most fragile part of the transfer. Where the effect is solid, it is a slab of ink with a plastic hand.

Halftoning converts opacity into dots: a 50%-opaque area becomes solid dots covering 50% of it, with bare shirt between. Each dot is full-strength ink with full underbase, so it presses and washes like the solid areas. Below the smallest dot that will bond, the fade stops cleanly. A real fade is a decision, because halftoning it, hardening it, removing it or flattening it to your shirt color each changes the look. A layer left at 98% opacity is not; that we snap to solid, because nobody can see the difference.

Export a PNG with real transparency, at least 300 times your print width in pixels, no matte color, and do not mirror it. For vector files, convert text to outlines first. The traps below are the ones we see most.

Every file gets a print-readiness check from our art team at no charge. We may fix the technical things that cause known DTF problems and do not change how the design looks. Anything that changes the design is your decision.

To have us make one of those changes, use Help me with my art when you add the print, in the studio or the cart. How a request works.