PDF to PDF/A Converter (with Readiness Check)

About the PDF to PDF/A Converter (with Readiness Check)

PDF/A is the PDF for keeping: everything a reader needs has to be inside the file (fonts, colour definitions), and anything whose meaning can change later — encryption, JavaScript, content fetched from elsewhere, multimedia — is not allowed. Courts, land registries, tax authorities and archives ask for it, and most refuse a file that only claims to be PDF/A.

This tool checks your PDF against PDF/A-2b and PDF/A-3b (the level that allows embedded files, as e-invoices need) and tells you exactly what stands in the way, with the clause each thing breaks. What a browser can put right it puts right: a PDF/A output intent (sRGB, or the colour profile a PDF/X file already carries), XMP metadata with the PDF/A identification, and the removal of JavaScript and the other forbidden actions, form-field scripts, XFA forms, multimedia annotations, transfer functions, print-exchange hints and presentation steps. What it cannot invent — an embedded font, a CMYK colour profile, an appearance for a comment — is listed as a blocker, and then the file is written without the PDF/A label, because a file that falsely claims PDF/A is worse than one that claims nothing.

Common uses

• Filing a document with a court, a registry or an authority that requires PDF/A.
• Finding out why a system rejected your “PDF/A” file.
• Preparing documents for long-term storage, so they still render in twenty years.
• Making an e-invoice PDF a proper PDF/A-3b container.
• Checking a batch of exports before someone else checks them for you.

What is checked, and what it is checked against

The rules are those of ISO 19005-2:2011 (PDF/A-2) and ISO 19005-3:2012 (PDF/A-3), conformance level B — “visual reproducibility”: the file must always look the same. Level A additionally needs full tagging for accessibility, which this tool does not attempt.

Clause numbers follow the published rule set of veraPDF, the open-source reference validator of the PDF Association: the rules for PDF/A-2 and PDF/A-3. Fonts and colours are checked against what is actually drawn: every page, form, pattern, Type 3 glyph, image and annotation appearance is read. Checked here:

• Fonts (6.2.11.4.1-1, 6.2.11.3.2-1, 6.2.11.3.3-1, 6.2.11.6-2/-3) — every font that draws visible text is embedded (a font used only for the invisible text of an OCR scan is exempt), symbol and TrueType fonts use the encodings PDF/A allows, CID fonts carry their glyph map, and non-standard CMaps are embedded.
• Colour (6.2.3, 6.2.4.3-2/-3/-4) — an output intent with a valid embedded ICC profile (a PDF/X file’s own profile is reused); DeviceRGB needs an RGB profile and DeviceCMYK a CMYK one, unless a DefaultRGB or DefaultCMYK colour space replaces them.
• File structure (6.1.2-1, 6.1.3-1/-2, 6.1.7.1-3, 6.1.7.2-1, 6.1.10-1, 6.2.2-2) — a PDF 1.x header, a file identifier, no encryption, no external stream data, no LZW (in streams or inline images), and resources that every content stream names itself.
• Graphics (6.2.5, 6.2.8, 6.2.9) — no transfer functions or halftone phases, no image alternates or OPI, no PostScript or reference XObjects.
• Annotations (6.3.1-1, 6.3.2, 6.3.3) — no multimedia, printable and visible flags, an appearance for every annotation that needs one.
• Forms, signatures and actions (6.4, 6.5, 6.1.12) — no XFA, no NeedAppearances, no scripts on form fields, pages or the document, only the action types PDF/A allows, and no signature that converting would break (usage rights a change voids are removed).
• Metadata (6.6.2.1, 6.6.2.3, 6.6.4) — an XMP stream with the PDF/A identification that agrees with the document information; metadata of a schema nobody describes is removed, everything else (such as the Factur-X data of an e-invoice) is kept.
• Embedded files (6.8) — forbidden at PDF/A-2 unless they are PDF/A themselves; allowed at PDF/A-3, with both names, a MIME type, a relationship and an /AF entry.
• Optional content, presentations, requirements (6.9, 6.10, 6.11) — named layer configurations with a complete layer order, no alternate presentations or presentation steps, no /Requirements.

Why the label is withheld

A PDF says it is PDF/A in its XMP metadata (pdfaid:part and pdfaid:conformance). Writing that in a file that does not conform is the one thing a converter must never do: the next system to check it rejects it, and whoever archived it believes it is safe. So while any blocker is left, this tool applies every fix it can, removes any PDF/A identification the file carried, and names the result …-pdfa-not-ready.pdf.

When nothing blocks it, the identification is written — and the report still asks you to confirm the file with veraPDF. These are the checks a browser can make, not a full validation: a validator reads the font programs themselves, the colour spaces of every image, the structure tree and much more.

What to do about the usual blockers

• Fonts not embedded — export again from the program that made the document with font embedding on. For a scan, run OCR PDF (its text layer uses an embedded font), or keep the pages as images.
• CMYK colours — convert the colours to RGB in the design program, or ask your print shop for a PDF/X file with their own profile. A CMYK output intent is a licensed printing profile that cannot be supplied here.
• Annotations without an appearance — flatten the comments and form fields into the pages with Flatten PDF, then convert the result.
• A form that asks the reader to draw its fields — open and re-save it in a PDF reader (which writes the appearances), or flatten it.
• LZW compression or external content — re-save the file from a current PDF writer; both come from very old producers.
• Encryption — remove the protection with Unlock PDF first; PDF/A files are never encrypted.

Limitations

• These are browser checks, not a validation: veraPDF reads much more (font programs, every image’s colour space, the structure tree) and has the last word.
• Level A (tagged PDF/A, for accessibility) is not attempted — only the B levels, which is what archives and courts usually ask for.
• The tool never converts colours, embeds fonts or rasterises pages: those change what the file is, and belong in the program that made it.
• The font programs themselves are not read (a glyph missing from an embedded subset, widths that disagree), and JPEG 2000 images, halftone dictionaries, blend modes, the validity of embedded ICC profiles and the implementation limits of clause 6.1.13 are not checked — veraPDF checks those.
• Metadata of a schema the file does not describe is removed to meet clause 6.6.2.3; the report lists each property that went.
• A password-protected or encrypted PDF cannot be read at all.
• A digitally signed file is not labelled PDF/A: converting rewrites the file, which breaks the signature, and PDF/A does not allow a broken one. Archive the signed original as it is, or convert the file before it is signed.

Privacy

Everything happens in your browser. What you enter or open here is not uploaded or stored by MySmartCoPilot.