Convert PDF to PDF/A

Three of the four requirements, and which one is missing

PDF/A is the archival profile — the one a court filing system, a journal submission portal or a national library asks for — and what it demands is that the file be self-contained, so that it renders identically in fifty years on software nobody has written yet. In practice that comes down to four things: an XMP metadata packet declaring which conformance level the file claims, an output intent naming the colour space its numbers are to be interpreted in, document information that agrees with the metadata, and every font used embedded in the file.

The first three are added here. The metadata packet is written with the conformance level you choose; the document information is rewritten to match it, because a validator will fail a file whose title says one thing in two places; and the output intent carries a genuine sRGB ICC profile, generated in the file rather than shipped as a binary blob — it is around four hundred bytes, which is small enough that there is no reason to fetch one.

The fourth cannot be done here, and this is the honest part. Embedding a font that is missing requires the font file, and the font file is precisely what is not in the document. Nothing in a browser can conjure it, and substituting a lookalike would produce a file that validates while the text reflows — worse than failing. The same applies to colour: converting a CMYK image to the declared output intent needs a colour management engine and the source profile. Most PDFs written this century by Word, LibreOffice, InDesign or a decent printer driver already embed their fonts, in which case what is added here is all that was missing.

So the right way to describe the result is PDF/A structure rather than a certified PDF/A file — a distinction that matters only because submission portals run a validator. Check yours before you rely on it. veraPDF is the reference implementation the PDF Association maintains, it is free, and it tells you precisely which clause failed instead of rejecting the file with no explanation.

When you want something else

Ghostscript is the tool that can do the part this cannot. gs -dPDFA=2 -dBATCH -dNOPAUSE -sColorConversionStrategy=UseDeviceIndependentColor -sDEVICE=pdfwrite -dPDFACompatibilityPolicy=1 -sOutputFile=out.pdf PDFA_def.ps in.pdf rebuilds the document through a full interpreter: it embeds fonts from the system where they are missing, converts colour through real ICC profiles, and refuses rather than silently producing an invalid file when it cannot comply. That is a genuinely different operation from adding the right metadata to a file that was already nearly conformant.

For a document that is scanned, PDF/A also wants a text layer to be archivable in any meaningful sense, and ocrmypdf --output-type pdfa in.pdf out.pdf does the OCR and the conversion in one step. It is free, and it is what most institutional workflows actually run.