Images → PDF

Bundle images into a PDF with predictable paper-size and layout behaviour. Built for the recurring "I need to send this batch to a printer" workflow.

When to use

• Preparing a photo set for a digital print shop (one-per-page on A4 / Letter).
• Building a proof sheet (multi-up).
• Sharing a document-style PDF (e.g. scanned receipts) without re-encoding.
• Bundling AI-generated images for client review.

Do not use this skill when:

• The user wants the PDF to also include text/markdown — that's a different workflow (typst / pandoc).
• The PDF needs OCR text layer — out of scope (run tesseract separately, then merge).
• A single image → PDF without paper-size constraints — img2pdf <input> -o <output>.pdf is enough; no skill needed.

Inputs

1. Inputs — directory, glob, or explicit file list. Required.
2. Output PDF path — required.
3. Paper size — A4 (default), A3, A5, Letter, Legal, Tabloid, or explicit <W>x<H><unit> (e.g. 210x297mm, 8.5x11in).
4. Mode — one of:
   • one-per-page (default) — one image per page, fitted to the paper with margins. Auto-rotates the page (portrait vs landscape) to match the image's aspect ratio.
   • multi-up — 2, 4, 6, or 9 images per page (--per-page N). Grid layout, each cell sized equally, image fitted within cell.
   • as-is — native pixel dimensions become the page size (use for digital archives or when DPI metadata is authoritative).
5. Margin — millimetres. Default 10mm (one-per-page), 5mm per cell (multi-up), 0 (as-is).
6. Sort order — filename (default, natural sort), exif-date (capture time from EXIF), mtime (filesystem mtime).
7. DPI — assumed DPI for fit calculations when source has no DPI metadata. Default 300 (print-quality).
8. Auto-orient — default on for one-per-page. Rotates page (not image) to match aspect — landscape image gets a landscape page. Disable with --no-auto-orient.

Procedure

1. Detect tool availability:
   • which img2pdf — preferred path.
   • which magick || which convert — fallback for cases img2pdf can't handle.
   • If neither exists, point at install-deps.
2. Enumerate inputs in the requested sort order (image extensions only: jpg, jpeg, png, webp, tiff, gif). Skip non-image files with a one-line note.

3. Mode: one-per-page with img2pdf — preferred path because img2pdf embeds JPEG bytes losslessly (no re-encode):

   img2pdf \
     --pagesize <paper-spec> \
     --imgsize <paper-minus-margins> \
     --fit into \
     --auto-orient \
     -o "<output>" \
     "<file1>" "<file2>" ...

   --fit into keeps aspect ratio, fits within page minus margins. --auto-orient rotates pages to match image aspect.

   img2pdf accepts JPEG, PNG, TIFF natively. For other formats (WebP, HEIC, AVIF), pre-convert to PNG via vipsthumbnail and stage in a temp directory; img2pdf will re-embed losslessly.

4. Mode: multi-up — img2pdf doesn't natively grid; build an intermediate composite per page using ImageMagick montage:

   # for each chunk of N files:
   montage <files> -tile <cols>x<rows> \
     -geometry <cell-w>x<cell-h>+<margin>+<margin> \
     -background white \
     "<staged>/page-<idx>.png"

   Then feed the staged page PNGs to img2pdf as one-per-page against the chosen paper size. Cell dimensions = (paper-w - 2*outer-margin) / cols - cell-margin and similar for height.

   Layout mapping: --per-page 2 = 1×2 portrait or 2×1 landscape (orient by paper); 4 = 2×2; 6 = 2×3 (portrait) / 3×2 (landscape); 9 = 3×3.

5. Mode: as-is — straight pass-through:

   img2pdf -o "<output>" "<file1>" "<file2>" ...

   No --pagesize or --imgsize; img2pdf uses each image's native dimensions and DPI.

6. ImageMagick fallback (only if img2pdf is missing or chokes on a format):

   convert <files> \
     -page <paper-spec> \
     -gravity center \
     -extent <paper-spec> \
     -density <dpi> \
     -compress jpeg -quality 90 \
     "<output>"

   This re-encodes JPEGs (quality loss), so prefer img2pdf when possible. Surface a warning when falling back.

7. Verify output: pdfinfo "<output>" (if available) or check file is non-empty PDF magic. Report page count.

Output

• The PDF at the resolved path.
• Summary: <n> images → <output> (<pages> pages, <size-MB>MB).
• Mode used + tool used (img2pdf primary / ImageMagick fallback).
• List of any skipped or staged-via-conversion files.

Notes

• one-per-page with auto-orient is the default because real-world prints want each photo at maximum size on the page regardless of original orientation. Disable for documents/contracts where every page must be uniform portrait.
• img2pdf is the right default: lossless JPEG embedding means the output PDF is essentially the JPEGs in a PDF wrapper — no quality loss, smallest possible file. ImageMagick re-encodes; only fall back when forced.
• For very large input sets (>200 images) ImageMagick montage becomes slow on multi-up mode — chunk and parallelise. For one-per-page mode, img2pdf handles thousands of pages without issue.
• Paper-size internals: A4 = 210×297mm, A3 = 297×420mm, A5 = 148×210mm, Letter = 8.5×11in (215.9×279.4mm), Legal = 8.5×14in, Tabloid = 11×17in.
• Print-shop-friendly defaults: A4, 10mm margin, 300 DPI assumption, auto-orient on. Override only with explicit reason.
• For mixed orientations in as-is mode without auto-orient, the resulting PDF will have heterogeneous page sizes — many printers handle this but some don't. Default to one-per-page with auto-orient for printer reliability.