Turn Scans and Photos Into a PDF

How to turn scans and photos into a PDF

A certificate photographed on a phone, a contract run through an office machine, a passport shot on a desk under a lamp. What you have is a dozen images of varying quality; what you need is one document an office or a portal will accept.

Before packing, the shots go through processing: they are straightened, rotated and brought into a mode suited to documents. Which mode you choose affects the result more than the quality of the photography does.

The processing profile decides almost everything

This is the main setting, and its default surprises people with colour scans.

The default is the text profile, and it removes colour. For a black and white contract that is exactly right: the text gets crisper, the background cleaner, the file several times lighter. But if you are scanning a document with a blue stamp or a coloured signature, the defaults will turn them grey.

That is not an error and not a loss of data in any legal sense, the stamp stays readable. But if the receiving side expects a colour scan and gets a grey one, questions follow. The mixed profile exists for those cases: it raises contrast and keeps the colour.

Straightening and rotation

Two settings are on by default, and both are almost always worth leaving on.

Straightening tries to correct a page shot at an angle. It is built for a small skew: a sheet laid slightly crooked on a desk, or a page that drifted in the feeder. If the photo was taken at a pronounced angle with perspective distortion, straightening helps little. Such a shot is easier to retake than to repair.

Automatic rotation works out the page orientation and brings it upright. It matters most for phone photography: the camera records orientation in the file properties, and without honouring that marker half the shots come out lying on their side.

There is one case for switching these off: when the processing got it wrong and turned a page the wrong way. Then it is simpler to disable the automation and rotate by hand with rotate-pdf.

Resolution and weight

Resolution is given in DPI, from 72 to 600, defaulting to 100.

The number here does not describe the quality of the photography but the resolution recorded in the finished PDF, that is what physical size the page comes out. The higher the DPI, the smaller the physical page for the same dot count, and the denser the text looks.

A hundred suits documents that will be read on screen. If the document is headed for print or for recognition, take 200 or 300: recognition accuracy on small type falls noticeably at low resolution.

Going up to six hundred is almost never necessary. That resolution makes sense for fine print work and archival digitisation; on an ordinary certificate it only inflates the file without adding legibility, because detail is limited by the quality of the shot, not by the number you wrote into the properties.

Step by step

1. Open scan-to-pdf and upload all the shots at once. 2. Make sure the files are selected in the sequence you want: that sequence becomes the page order, and you cannot change it inside the assembly. 3. Pick a profile to match the document type instead of leaving "text" on a colour scan. 4. Set the resolution according to what happens to the document next. 5. Run the assembly and page through the whole result, not just the first sheet.

A worked case: a document package for a portal

The typical job looks like this: you have to file an application with a passport, a certificate and a receipt attached, all shot on a phone at different times and lying in no particular order.

Order matters more than quality here, and there is a catch. Pages assemble in the order the files entered the upload, and you cannot rearrange them inside the scan assembly itself: there are no buttons and no dragging in that list. Meanwhile the phone names shots by the time they were taken, so if the certificate was reshot twice, the second frame will not sit next to the first.

Two workable paths follow. Either select the files in the right sequence at upload time, because the selection order is what becomes the page order. Or assemble it however it comes and then reorder the sheets with organize-pdf, where reordering exists. The second path is safer for a package of a dozen mixed shots: it is harder to go wrong reading thumbnails than filenames at the moment of selection.

If the shots are already in the right order and all you need is assembly, jpg-to-pdf closes the job faster: file order can be changed on the spot there, though it has no scan processing profiles.

Choose the profile by the most colourful document in the package rather than by the majority. If four of five sheets are black and white and the fifth carries a blue stamp, assemble the whole package in the mixed profile. The difference in weight will be small and the stamp will stay blue.

Shoot a passport spread whole, both pages in one frame, unless the requirements forbid it. Two separate shots of the left and right halves produce two sheets in the PDF, leaving the reviewer to assemble the spread in their head.

A finished package almost always runs into a weight limit. Assemble it in full first and compress afterwards: compressing one document works more precisely than processing each shot separately.

One more argument for the text profile where it fits: recognition works better on it. High contrast and grayscale strip out paper noise, colour casts from a lamp and the shadow of a fold, and it is exactly that debris recognition stumbles on. If the document is definitely headed through ocr-pdf, removing colour is preparation rather than loss.

What to do before uploading

Processing fixes skew and orientation. It does not fix the photography.

A shadow cast by your own hand or head over the sheet stays a shadow. The text profile will render it as a grey patch, and letters will disappear into it. Shoot with the light source to the side rather than behind you.

Cropping is not automatic either. If the shot shows the desk as well as the sheet, the desk goes into the PDF along with the document. Surplus can be trimmed with crop-pdf after assembly, but framing tighter is simpler.

No profile rescues a blurred frame. Checking is easy: if small type is unreadable when you zoom in on the phone screen, it will be unreadable in the PDF too.

What comes next

An assembled document usually passes through another step or two.

If the file came out heavy and you would rather not change the profile, compress-pdf takes the weight off. If search has to work across the document, you need ocr-pdf: it recognises the text on the images and adds it as an invisible layer, leaving the picture as it is. If there are many scans belonging to different matters, split-pdf helps sort them into separate documents.