Preflight Check
Read the profile, evidence coverage and findings before choosing a correction or requesting production review.
Preflight records the checks made on a PDF before printing. Use it to locate risks, request a corrected source and discuss the file with your printer. The report describes the selected profile and actual evidence; it is not a general certificate that the file will print correctly.
Run a check
- Open Preflight, choose the screening profile and select your PDF.
- Browser processing keeps the PDF on your device. Its document limit is 200 MB and 500 pages; MB means 1024 × 1024 bytes. Complexity, memory and time budgets can still refuse a smaller file.
- Read the profile, measurement source, evaluated and undecided rule counts, findings and limitations.
- Download the report. If you use report sharing, read who can open the link.
- Fix source-design issues in the authoring application, export a new PDF and check that version.
A browser failure does not authorize an automatic upload to a server. Choose an explicitly available server flow and its required access/entitlement when you want server processing. See server runs for that separate flow.
What to inspect
| Area | Read the evidence and the job requirement |
|---|---|
| Page geometry | Compare the finished size, orientation and page boxes with the order. MediaBox describes the PDF page extent, not the printing sheet. |
| Bleed and safe area | Check every cut edge and the required allowance. A declared BleedBox does not create missing background artwork. |
| Images | Use effective PPI at the intended placed size. A profile threshold is not a universal 200 or 300 DPI rule. |
| Fonts | Review embedding, glyph and font-stream warnings. Correct missing content at source; a subset flag alone is not proof all glyphs are present. |
| Colour | Check colour spaces, output intent, transparency and named colourants against the intended RIP/DFE. RGB may be valid in a suitable PDF/X-4 workflow. |
| Ink | Read the selected paper/profile limit and the measurement type. A page average, a maximum and an estimate answer different questions. |
| Security | Review active content, unsupported actions and incomplete traversal. No findings is not malware certification. |
| Page count | Follow the selected binding and imposition method. A multiple of four is not a requirement for every book or print job. |
See image resolution, bleed, fonts and PDF requirements.
Read the result in its context
A single-file diagnostic, a browser batch classification, a workflow check and a production decision are different results. A diagnostic without canonical decision evidence remains review. Batch Ready (checked in the browser) additionally requires trusted, complete and verified local evidence. Workflow clean requires no error/warning, at least one evaluated rule, no undecided rule, complete analysis and trusted structure; information notes alone do not block that classification.
Commercial-print and book workflows have mapped engine profiles. Packaging currently checks the baseline and discloses the specialist checks that did not run. Unmapped profiles are not silently accepted. Server coverage depends on its enabled profile configuration and available engine. See reading results.
A failed, cancelled, timed-out or unsupported check is not a pass. Sampled pages do not certify uninspected pages. Read large-file scope and ask the printer for a broader inspection when the actual evidence does not cover the production risk.
PDF/X and production authority
If an older report uses internalPrintRiskScreeningV1, read it as structural risk screening,
not independent standards conformance or a printing certificate. The older
signatureStructureInventory label means a signature-structure inventory, not cryptographic
signature, certificate-chain or current-trust validation. Use the report's actual profile ID,
version and collected evidence when deciding what it establishes.
The public PDF/X-1a and PDF/X-4 choices are structural screening profiles, not independent conformance validators. Accessibility declarations similarly do not establish PDF/UA compliance. A profile name, successful parse or high confidence cannot replace applicable validator evidence.
A production decision requires the native canonical report and a fresh verified decision receipt bound to that exact source, report, profile and rule registry. An expired, differently bound or missing receipt sends the job to review. A report's customer/supplier presentation, a download, customer approval, or a visible preview cannot grant that authority.
Original and derived files
Preflight itself does not change the source PDF. NowToPrint does not automatically force them into CMYK when RGB or mixed spaces are detected. RGB/spot to CMYK conversion, transparency flattening, trapping generation and font conversion/outlining remain external-only corrections when the target workflow actually needs them. Agreement to print does not authorize these transformations.
Supported page organization is a separate operation. Keep the original, read the specific changes and output receipt, and inspect/recheck the derived PDF. A page-box change cannot manufacture missing bleed artwork; successful byte output does not establish production proof.
Reading advanced evidence
The internal ntp-pdf-source foundation provides bounded, content-addressed source access.
A digest binds raw PDF bytes; a retained handle and directory-sync capability describe separate
filesystem guarantees. If a required capability is unsupported, do not treat it as enforced.
This foundation alone does not make a preview or preflight result production-ready.
The ntp-pdf-syntax and ntp-pdf-parser foundations supply syntax and indexed lazy parsing;
these internal facts are not production-ready preflight or a conformance result.
A capability truth map distinguishes measured, estimated, heuristic and unsupported findings. When Rust native measurements are supplied, their identity and limitations stay attached. A browser preview is not measured PDFium raster evidence; even measured raster evidence is not RIP output, colour-separation proof or a certified press proof.
Packaging evidence may include barcode quiet-zone, contour bleed, separation proof sheet requirements and supplier finishing constraints when the necessary measurements are present. An absent specialist measurement remains unmeasured; a baseline packaging check does not perform all these checks or redraw the artwork.
A public trustPassport projection, when returned by the selected API flow, explains its evidence
and limitations. Any rework or reject estimate is a deterministic policy estimate for review;
it is not measured print quality or automatic production permission.
Benchmark claims
Synthetic examples, licensed public files and private production evidence are separate evidence sets. A local fixture result does not establish a real-world success rate or parity claim.
Outcome dashboard
An internal outcome dashboard may compare labelled outcomes, reuploads and disputes with policy estimates. Those operating metrics do not certify this PDF or replace its exact-file evidence.
If you need a correction
For a low-resolution image, find a better source or reduce its placed size. For missing bleed, extend the artwork at source to the printer's required allowance. For missing fonts, export with the necessary fonts embedded where permitted or request specialist review. For uncertain RGB or transparency handling, agree the output intent and RIP/DFE plan; do not apply a generic destructive conversion. Keep failed or unsupported results visible in the handoff.
Was this article helpful?