Merge and split PDFs from the command line

pdfunite a.pdf b.pdf out.pdf joins documents and pdfseparate in.pdf page-%d.pdf breaks one apart. Both come from poppler-utils and are fast and lossless. What they can't do — extract a page range as a single document, or compress — is ghostscript's job, and that split of labour is the whole article.

1. How many pages is this

$ clai how many pages does merged.pdf have→ pdfinfo merged.pdf | grep '^Pages'Pages:           4

pdfinfo prints a couple of dozen metadata lines and you usually want one of them. It doubles as the sanity check after every operation below.

2. Join several files

$ clai merge invoice.pdf and contract.pdf into one document→ pdfunite invoice.pdf contract.pdf merged.pdf

The last argument is always the result, everything before it is a source, in the order they'll appear. The command prints nothing — check with pdfinfo.

3. Break it into single pages

$ clai split merged.pdf into one file per page→ pdfseparate merged.pdf page-%d.pdfpage-1.pdfpage-2.pdfpage-3.pdfpage-4.pdf

The %d is required — the page number goes there. The listing above comes from ls -1; pdfseparate itself is silent.

4. Extract a page range

$ clai pull out pages 2 through 3→ pdfseparate -f 2 -l 3 merged.pdf slice-%d.pdfslice-2.pdfslice-3.pdf

-f is the first page, -l the last. Note what you got: two files, not one two-page document — and the numbering keeps the original page numbers.

5. A range as one document

$ clai extract pages 2 to 3 into a single document→ gs -q -sDEVICE=pdfwrite -dNOPAUSE -dBATCH -dFirstPage=2 -dLastPage=3 -sOutputFile=pages-2-3.pdf merged.pdf

poppler has no command for this, so ghostscript takes over. Verified: pdfinfo pages-2-3.pdf reports Pages: 2. Without gs the route is pdfseparate followed by pdfunite slice-2.pdf slice-3.pdf range.pdf, which gives the same two pages.

6. Make the file smaller

$ clai compress the pdf so it fits in an email attachment→ gs -q -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -dNOPAUSE -dBATCH -sOutputFile=small.pdf merged.pdf

Verified on the demo file: 84991 bytes before, 31876 after — nearly a third of the size. /ebook is the middle profile; /screen squeezes harder and shows it on images, /prepress protects quality. Text is untouched either way; what shrinks is the raster images inside.

Gotchas

• pdfunite overwrites its last argument. That slot is always the output. pdfunite part1.pdf part2.pdf with no third name destroys part2.pdf — there is no "last file is a source" convention here.
• pdfseparate never joins. Even with -f and -l you get a pile of one-page files. Putting them back together is pdfunite's job, or ghostscript's.
• Neither tool opens an encrypted PDF. Remove the password first (qpdf --decrypt where it's available); otherwise both simply refuse to read the file.

Related questions

How do I rotate pages? Ghostscript can re-emit a file with -dAutoRotatePages=/None, but rotating one specific page is far simpler in qpdf.

How do I get the text out? pdftotext file.pdf - prints the text to the terminal. For scans with no text layer you need OCR instead.

How do I turn pages into images? pdftoppm -png -r 150 file.pdf page writes page-1.png, page-2.png and so on at 150 dpi.

See also

• Batch resize images
• Split a large file into parts, and put it back together
• Convert and compress video with ffmpeg

CliAI knows which of the two toolchains does the job you described, and shows the command before it runs. Install it in one line.