Turning pictures into a PDF feels like it should be the simplest thing a PDF tool does. Put image in, get page out. And it is simple — right up to the moment someone asks why the pages are a strange size, or why the file is bigger than the pictures were, or why the text in the scan cannot be searched.
All three have the same answer, and it fits in one sentence: a PDF page is measured in points, an image is measured in pixels, and nothing automatically converts between them.
Pixels are not a size
An image 1600 pixels wide has no physical size. It is 1600 samples of colour. How big it should be printed or displayed is a separate question, and the answer is a resolution: at 300 DPI those 1600 pixels are 5.3 inches; at 72 DPI they are 22.2 inches. Same image, same pixels, four times the paper.
A PDF page, by contrast, is measured in points, and a point is exactly 1/72 of an inch. A4 is 595 × 842 points. US Letter is 612 × 792.
So when an image becomes a page, something has to decide how many points wide that image should be, and there are two reasonable answers. JPG to PDF offers both.
Original image size maps one pixel to one point. We measured it: a 1600 × 1200 pixel image became a page of exactly 1600 × 1200 points — 22.2 inches wide, roughly A1. Not a mistake, just the arithmetic, and it is the right default for anything that will be read on a screen because no white border is added around the picture.
It explains the two things people notice, though. On screen it looks perfectly normal, because a screen renders a page at whatever zoom fits the window. Sent to a printer it is enormous, and the printer will either scale it down or spread it across several sheets.
A4, Letter or Legal is the other answer: the page is that paper size and the image is scaled to fit inside it, never cropped and never stretched. The same photograph placed on A4 gave a 842 × 595 pt page — landscape, chosen automatically, because the image was wider than it was tall. That is the figure above, and it is the setting to use when the document is going to be printed.
What the file does depends on how the image was encoded
We took one photograph, saved it two ways, and put both through unchanged.
- As a JPEG: 57 KB in, 58 KB out. Essentially untouched — the JPEG is embedded in the PDF as it already was.
- As a PNG: 196 KB in, 29 KB out. Converted to JPEG on the way in.
Same picture, same pixels, very different arithmetic — and the difference is not the PDF, it is the source encoding. PNG is lossless, which is excellent for screenshots and line art and wasteful for photographs; converting it to JPEG throws away detail the eye does not miss, and the file drops sharply.
The practical consequence is a good one: a JPEG that goes into a PDF is not re-encoded, so it does not lose anything. That was not always true, and it is worth knowing if you care about generation loss — the old advice to avoid repeated round trips applies to the conversions that re-encode, not to this one.
Going back out with PDF to JPG is a fresh encode, so the usual rule still holds there: each export is a new JPEG, and JPEG is lossy every time. Once is invisible; five round trips is not, and the damage never comes back.
A picture of text is not text
This catches people out constantly, and it is worth being blunt about.
When a photograph of a document becomes a PDF page, the PDF contains a photograph. It does not contain the words. Ctrl+F finds nothing, you cannot select a sentence to quote, and a screen reader gets silence. The page looks like a document and behaves like a picture, because it is one.
Adding a text layer is a separate operation: OCR recognises the characters in the image and writes them behind it, which is what makes the page searchable. It is a recognition pass rather than the original text, so it is a best guess — but it is the difference between a findable document and a flat image.
If you are converting photographs of paper rather than photographs of things, assume you want OCR afterwards.
Practical consequences
Order is set before you convert, not after. Images go onto pages in the order shown on the page, which is why they are draggable. Filenames sort in ways that surprise — IMG_9.jpg lands after IMG_10.jpg in a plain alphabetical sort — so check the sequence rather than assuming the camera’s numbering survived.
Shrink before you convert, not after. If the pictures are coming off a modern phone at 12 megapixels and the document is for email, resizing the images first gives a far better result than compressing the PDF afterwards. Compression works by reducing image resolution, so you end up at the same place having re-encoded everything twice.
Mixed orientations are a setting, not a problem. At Original image size, portrait and landscape photographs produce portrait and landscape pages — correct, and untidy. Choosing a paper size fixes it: with orientation on Auto each sheet turns to match its image, and setting it to Portrait forces every page the same way regardless of what is on it.
Screenshots are the opposite case. A screenshot is captured at screen resolution, roughly 72 to 96 pixels per inch. One pixel to one point renders it at close to its natural size and it looks right. The same screenshot forced onto A4 and enlarged looks soft, because there were never enough pixels to fill that much paper.
The rule of thumb
Decide what the document is for before you convert, because the answer changes what matters.
For reading on screen, leave the page size on Original and the question never arises. For printing, choose the paper size you are printing on and work backwards from it: 300 DPI on A4 needs an image about 2,480 pixels wide, and no tool can add pixels that were never captured. For archiving or searching, convert, then run OCR, and check that selecting a line of text actually highlights it.
You can put your own images through JPG to PDF and open the result’s page size in any reader — the arithmetic above will be sitting right there.