JPG to PDF

The pictures are embedded, not re-encoded

A JPEG is stored inside the PDF exactly as the camera wrote it: the file records "this page holds JPEG data" and hands the original bytes to the reader, so not a single pixel is touched and there is no second generation of quality loss. A PNG is stored losslessly, which for a screenshot or a logo means the picture arrives with its edges intact.

Because nothing is re-encoded, the PDF comes out roughly the size of the pictures you add: a 4 MB phone photo still costs about 4 MB. That is a deliberate trade, because the alternative is discarding detail, and that is not a decision this tool should make for you. Compression is a separate job with its own page.

One picture is one page. Pages follow the order you add the files, or alphabetical order by file name if you prefer it, and nothing is deduplicated - add the same picture twice and you get two pages, because printing a page twice is sometimes the point.

Sheets, resolution and the phone-photo trap

A page can be the size of the picture or a standard sheet. "Page matches the image size" puts each picture on its own page at a resolution you choose: 96 dots per inch is the screen convention, where one pixel becomes 0.75 points, and 72 dpi keeps one pixel to one point so nothing is resampled. Standard sheets use real A4 or US Letter dimensions, and each picture is centered inside the margin you set without being distorted - a wide photo on a portrait sheet is scaled until one side fits, never stretched.

The trap is Exif. Phones store the sensor read-out as it came off the chip, often landscape, and record "turn this 90 degrees" in the image metadata instead of moving pixels. Software that only reads pixels shows the picture on its side, which is why so many converted PDFs arrive sideways. The orientation flag is read here and the picture is turned while it is placed on the page, so the page itself is written upright - it does not depend on a rotation instruction that a printer or an upload system might ignore.

Four of the eight Exif values also describe a mirror. A mirror cannot be expressed by rotating a picture, so those pages come out upright but not flipped, and the result says so rather than quietly handing you something wrong.

What a picture cannot give you

A JPG of a page of text is a picture of text. No amount of careful embedding turns those pixels into characters: the output is not searchable, its words cannot be selected or copied, and a screen reader sees an image. Turning pixels into text is OCR, which is a different job - if a document has to be searchable, recognize the text first and convert afterwards.

Two more limits are worth knowing before you start. The first is size: without re-encoding, a stack of 8 MB photos makes an 8 MB PDF. The second is enlargement: a small picture stretched onto an A4 sheet is interpolated on the page and can look soft, so the result reports how far each picture was scaled and points you at "page matches the image size" if that matters.

Order, naming and the limits that cut the stack down

Pages follow the order you add the files, or alphabetical order by file name if you pick that instead - and because that comparison is a real alphabetical one, shot-2 sorts before shot-10 the way folders already do. Nothing is deduplicated, so adding the same picture twice gives two pages, which is how a page that has to appear twice is made in the same pass as the rest.

One picture makes one page, and a job takes at most 20 pictures of up to 25 MB each, all read in this tab and never uploaded. The output keeps the name of the first picture unless you type one. The pictures are embedded as they are - JPEG data stays JPEG data and nothing is re-compressed - so the PDF holds exactly what you shot.