What Print Prepper does: bleed, crop marks and a print-ready PDF

Customer files turn up wrong in one of five ways: too small, no bleed, wrong shape, wrong details printed on the artwork, or wrong colour. Usually two or three of them at once, on the same file.

Print Prepper deals with four of those. On the fifth it has an answer you’d want to hear before you sign up. Here’s each one, what the software does with it, and where it stops.

Too small

You upload PDF, Illustrator .ai, TIFF, JPEG or PNG, and you set the DPI you want out the other end.

What happens next depends on what the file is. Photographs can go through Real-ESRGAN. Line art and logos never do. They take a Lanczos resample, because a model rounds off the edges you want kept hard.

The conditions on the photo path are not the ones most people would guess. Real-ESRGAN runs only when the scale factor is 1.2 or more, and only on sources up to two million pixels. So the bigger the customer’s photo, the less likely it is to take the AI path. Real-ESRGAN also only offers 2× and 4×, so whatever factor you asked for gets snapped up to the next one of those. After the resample there’s a sharpening control: off, or unsharp mask at sigma 0.5, 1.0, 1.5 or 2.0.

The photo-or-graphic decision gets made once, on page one’s raster, and then applied to the whole job. It comes out of a thousand-bucket colour histogram with a four per cent threshold, and there’s no model involved in that call. If the check errors, it falls back to treating the file as a photo. A twelve-page PDF with a photo cover and eleven pages of type gets classified off that cover, so a mixed document is the one to watch.

Nothing here guarantees you 300 DPI. You set a target and the tool tells you what it got.

No bleed

Set the bleed in whole millimetres and the artwork is mirrored outward past the trim on all four edges.

The pixels along the edge get flipped and repeated into the bleed band. On anything with recognisable content near the trim you will see it happen: a reversed half of a face, a line of type reflected back on itself. On texture, sky and flat colour it’s invisible and does the job perfectly well. Look at the proof before you commit to a full bleed on a busy edge.

Then the crop marks. Eight strokes, one pair at each corner, drawn on every page. The length, the offsets (18pt and 6pt), the width (0.25pt) and the colour (black, DeviceRGB) are constants in the code, so you can’t change any of them. A TrimBox is written into the PDF; there is no BleedBox and no ArtBox. There are no registration marks, no colour bars and no slug information. Anything past that has to come from somewhere else in your workflow.

Wrong shape

Trim comes from a preset or from typed millimetres, plus portrait or landscape. The presets cover A0 through A7, US Letter, Legal and Tabloid, postcards at 4×6, 5×7 and 6×9, the 99×210 DL flyer, business cards at 90×55 and 89×51, square cards and square formats, photo prints, and 18×24 and 24×36 posters.

When the artwork is the wrong proportion for the trim, there’s padding. Padding extends the canvas before the crop, so you have something to frame against instead of cutting into the design.

Two things to know about it. The padding value is a percentage, a whole number from 1 to 100, taken from the source image’s own dimension. It isn’t millimetres and it isn’t bleed. And applying it a second time stacks another bar on top of the first instead of replacing what’s already there. Every step is kept as a revision, so you can step back through ten of them, redo, or restore a version you had abandoned.

Padding is set per page. A multi-page source is split into individual pages on upload, and each page then gets cropped, padded and edited on its own. They come back together as one PDF in the original page order.

Wrong details

A wrong phone number isn’t a prepress problem, but it lands on the same desk anyway. Print Prepper will edit the artwork itself.

It works one page at a time, and only while the job is at the crop stage. A vision model reads the page and suggests up to four concrete edits it can see are needed: change the phone number to 0400 000 000, remove the second email address under the logo, that sort of thing. You then type a free-text instruction. The only other control is an aspect-ratio hint. There’s no masking, no brush and no region select.

Generation takes roughly 60 to 180 seconds. What comes back is a proposal, and nothing lands on the page until you say so. Your options are Apply, Discard or Refine. The image before the edit is snapshotted, so one click puts it back the way it was.

Because the model picks its own output size, the app measures the returned pixel width against your trim width and reports the DPI you actually ended up with against the DPI you asked for.

Attempts are capped at 200 per organisation per month by default. A generation that fails still counts, since it cost the same call as a good one, but discarding an attempt deletes it and gives you that one back.

Wrong colour

Print Prepper doesn’t fix this one.

The pipeline is RGB end to end, with no ICC profile embedded and no CMYK conversion happening anywhere in it. If a source arrives in CMYK or carries spot colours, the preflight step warns you that the information will be lost, and that warning is the extent of the colour handling.

Conversion stays where it already sits, with your RIP and your press profiles. If that rules the tool out for the work you do, you’re better off finding that out in the first hour rather than at the first proof.

What comes out

Two PDFs. One press-ready multi-page file for the printer, and one low-resolution soft proof to send the client for sign-off, with PROOF across every page so it can’t be mistaken for the press file.

Around that there’s an organisation you invite the rest of the shop into by email, with authenticator-app two-factor. Everyone in it can run jobs; the roles cover who administers the account.