What makes a photo look like a photo
Three things give it away. The lighting is uneven, so one side of the page is bright and the other sits in shadow. The paper is grey rather than white, because the camera exposed for the whole scene and paper is the brightest thing in it. And the text is grey rather than black for the same reason.
A flatbed scanner avoids all three by lighting the page evenly from a fixed distance with a known light source. It never has to guess at exposure because the conditions never change.
A phone has to guess, and it guesses for the whole frame rather than for the page. Everything that follows is either about giving it a better scene to guess from, or correcting the guess afterwards.
Lighting is the thing to fix first
The most common ruin is your own shadow falling across the page, which happens whenever the light is behind you. Move so the light comes from the side or from behind the page, not over your shoulder.
Diffuse light beats direct light. An overcast window is close to ideal. Direct sunlight creates a hard bright patch and a hard shadow edge that no processing recovers cleanly, and a single overhead bulb creates a bright centre and dark corners.
Two lamps at roughly 45 degrees from either side is the studio version if you scan often. For occasional use, a window and a bit of positioning is enough.
Angle and framing
Hold the camera parallel to the page, directly above the centre. Shooting at an angle makes the far edge smaller than the near one, and text at the far end stays smaller and harder to read no matter what processing you apply.
Get the whole page in frame with a small margin around it. Cropping tight to the edges risks losing a corner, and a margin gives you room to straighten afterwards.
Tap to focus on the text before shooting rather than trusting autofocus to pick the page. A soft photo cannot be sharpened into a legible one, and this is the single most common reason a scan is unusable.
What processing can fix
Uneven brightness across the page is correctable. Local contrast normalisation removes the broad brightness variation, so a shadow across one half stops mattering, and then the tonal range is stretched so paper goes to near-white and ink to near-black.
The result reads cleanly on screen, prints properly, and compresses to a small fraction of the original photo's size — a mostly-white image with black text is the easiest thing there is to compress.
What processing cannot fix is geometry, focus, or missing detail. If the corner was out of frame it stays out of frame, and if the text was blurred it stays blurred.
Colour, greyscale, or black and white
For plain text, high-contrast black and white is usually the best result: cleanest to read, smallest file, and closest to what a scanner produces.
Greyscale is the right choice where the document has photographs, signatures in pencil, or shading that carries meaning. Pushing those to pure black and white loses information.
Keep colour when it is load-bearing — a stamped seal, a coloured signature, a highlighted section, a chart. Colour files are larger, so only pay that cost when the colour is doing something.
Multi-page documents
Shoot every page in the same position with the same lighting before you process any of them. Consistency across pages is more noticeable to a reader than the absolute quality of any single one.
Process each page identically, then combine them into a PDF at the end. Mixing settings between pages produces a document that looks assembled from different sources, which undermines it if the document is meant to look official.
Name files in page order as you go. Re-sorting a twenty-page scan afterwards is a genuinely miserable job.
What a scan does not give you
A processed image of a page is not searchable or editable text. Extracting the words requires optical character recognition, which is a separate operation from cleaning up the image.
If someone needs to copy text out of the document, or you need to search a large archive, plan for OCR as an additional step rather than expecting a scan to provide it.
For most everyday purposes — sending a signed form, submitting a receipt, filing a letter — a clean image is exactly what is wanted and OCR is unnecessary.
Privacy is worth a thought
The documents people scan are contracts, medical letters, bank statements, passports, and payslips. Those are exactly the files worth being careful about where they go.
Many scanning apps upload to a server for processing, and some are free specifically because of what they learn from that. Read what a tool does before feeding it a passport.
Processing that runs entirely on your own device sidesteps the question. There is no upload, no retention policy to trust, and nothing to delete afterwards.
Frequently asked questions
How do I make a phone photo look like a scan?
Shoot parallel to the page with even, diffuse light and no shadow across it, then process to flatten lighting and push paper to white and ink to black.
Why is my scan grey instead of white?
The camera exposed for the whole scene rather than the page. Processing that normalises local contrast and stretches the tonal range fixes it.
Does scanning give me editable text?
No. A scan is an image of the page. Extracting text requires optical character recognition, which is a separate step.
Should I scan in colour or black and white?
Black and white for plain text — cleanest and smallest. Greyscale where shading matters. Colour only when the colour carries meaning, like a seal or a chart.
How do I make a multi-page PDF?
Shoot every page in the same position and light, process them identically, then combine the images in a PDF tool at the end.
Is it safe to scan an ID document with an online tool?
Only if it processes on your own device. Many upload to a server, which is a real consideration for passports, statements, and medical letters.