Create a PDF from Photos

The document exists only as a phone photo, and it has to go out as a single file. Assembly takes a minute, but the result depends heavily on two settings that do not default to what most people expect. Here is what actually happens to a shot on its way into a PDF.

The page fits the photo, not the other way round

The page size setting defaults to source. That is not A4. Each sheet is built around the individual shot: its own pixel dimensions are taken, and margins are added on all four sides. A 4032 x 3024 photo produces a landscape sheet; the next shot from a different camera produces a sheet of a different size.

On screen the mismatch is barely visible, because the reader scales each page to the window. It shows up in two places. When printing, the driver fits every non-standard sheet to the paper in its own way, so the margins come out uneven. And intake forms that check the sheet format reject a document made of odd-sized pages without explaining why.

So the rule is short: if the file is going to a person by email, source is the friendlier choice, because the shot is never rescaled. If the file is going to a printer or a form, switch to A4 or Letter.

The canvas for those two formats is built on a fixed grid: 1240 x 1754 pixels for A4 and 1275 x 1650 for Letter, which corresponds to 150 pixels per inch. The output resolution value in DPI does not change the physical size of such a sheet; it only applies in source mode, where the canvas is built directly around the photo's own pixels.

Three fit modes and what each does to the shot

For documents, only Contain is safe. Cover trims along the edges of the sheet, and the edges of a document are exactly where the reference number, the date and the signature usually live.

A4 carries a separate trap: the sheet is always portrait and cannot be turned. In Contain mode a landscape shot lands as a narrow strip in the middle, with wide white bands above and below. For landscape frames, either stay on source or crop the photo to portrait proportions before uploading.

How to build the PDF

1. Open jpg-to-pdf and upload every shot at once. 2. Check the order in the upload list: it becomes the page order. 3. Pick the page size. source for email, A4 or Letter for printing and forms. 4. Leave the mode on Contain if the frame holds a document. 5. Run the build and download the file. 6. Open the result and page through to the end; a missing or sideways sheet is obvious immediately.

You set the page order, not the capture date

Pages come out in exactly the order the files sit in the upload list. Capture date, file name and the order you tapped them in the gallery have no effect.

There are two ways to change it. On a computer you drag a file with the mouse. On a phone dragging will not work: the list uses the plain HTML drag-and-drop mechanism, and that does not respond to a finger. For phones there are up and down arrows next to each file, and they are the only working route.

The file's position number sits to the left of its name, which makes it easy to check the sequence without opening any preview.

Why some shots stand upright and others land sideways

A phone almost never rotates the pixels when it shoots. It stores the frame as captured and writes a tag into the EXIF data saying "turn this one 90 degrees". The auto-rotate switch is on by default, and that tag is what it acts on.

So the shots that land sideways are not random: they are the ones with no tag. It gets lost when the photo has been re-saved by an editor, arrived through a messenger that recompressed it, or is a screenshot. Such a file looks no different from the rest, and then appears lying on its side.

The cure comes after assembly: open rotate-pdf, mark only the affected pages and turn them. There is no need to spin the whole document, since rotation applies to the sheets you select.

Margins and the weight of the finished file

The page margin defaults to 12 points, roughly 4 millimetres. Zero puts the shot flush against the edge of the sheet, which looks good in a photo album and badly in a document headed for a printer: most printers have a non-printable border, and part of the content ends up inside it.

The finished PDF weighs about as much as the source shots added together, because the tool does not recompress the images or lower their resolution. A modern phone produces 3 to 5 MB per frame, so ten pages easily reach 40 MB. If the file has to travel by email or through a form, run it through compress-pdf; photographs are where compression pays off most visibly.

What the tool deliberately does not do

Do not expect more than it offers. jpg-to-pdf does not straighten a skewed page, does not raise contrast and does not remove the grey of the desk behind the sheet. It takes the pixels as they are and places them on a page. That split is deliberate: scan processing lives in scan-to-pdf, which has real deskewing and the Text, Mixed and Photo profiles.

Three more things are worth knowing in advance.

From a multi-frame file only the first frame is taken. An iPhone Live Photo saved as a HEIC sequence and an animated GIF both become a single page holding the first frame.

PNG transparency is not preserved; it is filled with white. For a document that is the right behaviour, but a logo on a transparent background gets a white rectangle around it.

Oversized images are rejected outright: the ceiling is 150 million pixels per shot. An ordinary camera does not come close, but a stitched panorama or an export from a graphics editor can.

What to do with the finished file next

A photograph of a document is still a picture. You can read the text, but you cannot find it by search or copy it, and a form that checks for a text layer will not accept the file. Recognition is a separate step through ocr-pdf: the letters stay where they are and an invisible text layer is laid over them.

If the desk, the edge of a keyboard or your fingers made it into the frame, trim them off with crop-pdf. The crop is set visually on the page, and the values are stored in points, millimetres or inches, whichever you pick.

Keep the operations in this order: assemble, rotate, crop, recognise, compress. Compression goes last, because any later step inflates the file again and the work has to be repeated.