Every problem with a scanned document was created at the moment of scanning. Nothing downstream fully recovers from a bad capture.
The three settings that decide everything
| Document | DPI | Colour mode | Rough size/page |
|---|---|---|---|
| Text, for reading | 200 | Greyscale | 100–200 KB |
| Text, for OCR | 300 | Greyscale | 200–400 KB |
| Text with a colour logo | 300 | Colour | 600 KB–1 MB |
| Photographs | 300 | Colour | 1–3 MB |
| Fine print, small type | 400–600 | Greyscale | 500 KB–1.5 MB |
| Archival, legal evidence | 600 | Colour | 3–8 MB |
Black-and-white mode — one bit per pixel, no grey at all — is worth avoiding despite the tiny files. It destroys anti-aliasing on text edges and turns any faint pencil, stamp or signature into either solid black or nothing. Greyscale costs three times as much and keeps everything.
Multiple pages into one file
This is a scanner setting rather than a post-processing job, and it is worth finding once.
A document feeder scans a stack into one PDF by default. On a flatbed, look for "Add page" or "Scan more pages" in the dialog before saving — most software offers it and most people miss it, then end up with forty separate files.
If you already have the separate files, Merge PDF combines them, and the only real trap is ordering: files sort alphabetically, so page10 lands before page2 unless the numbers are zero-padded. That and the rest of the merge failure modes are in how to merge PDF files.
Phone scanning
A modern phone scanning app is genuinely competitive with a flatbed for text documents, because the apps do perspective correction, cropping and contrast enhancement automatically. Both iOS Notes and Google Drive include one at no cost.
What a phone cannot match is even illumination. A flatbed lights the page identically everywhere; a phone captures whatever the room is doing, which is why phone scans often have a bright patch and a dark corner.
- Indirect daylight beats any indoor light. Near a window, not in direct sun.
- Never use the flash. It produces a hot spot in the middle and shadows at the edges.
- Shoot straight down. Perspective correction works, and works better on less distortion.
- Watch your own shadow. Leaning over the page puts you between it and the light.
- Use a plain contrasting surface. Edge detection needs to find the paper.
Straightening and sharpening after the fact
Both are possible and both cost something, which is the argument for getting the capture right.
Straightening a skewed scan means re-rendering the image at an angle. Every pixel is interpolated, so text softens slightly. A degree or two is invisible; ten degrees is noticeable.
Sharpening does not add detail — it increases local contrast at edges so the eye reads it as sharper. Overdone, it produces halos around every letter and looks worse than the original.
Rescanning takes two minutes and costs nothing. Correcting takes longer and always loses a little. If the document still exists on paper, rescan it — and note that page rotation cannot fix skew at all, since it only accepts multiples of 90, which is covered in how to rotate a PDF so it stays rotated.
Handwriting
Scanning handwriting is easy; converting it to text is not.
Standard OCR is trained on printed type. On neat handwriting it manages perhaps 60%, and on ordinary handwriting considerably less. Specialist handwriting recognition exists and is better, but nothing yet approaches the reliability people expect from printed-text OCR.
Scan handwritten notes at 300 DPI greyscale and treat them as images. If you need them searchable, the honest options are to type them up, or to accept that OCR output will need line-by-line correction — which for anything longer than a page is slower than typing.
Size, after the fact
A 20-page scan at 600 DPI colour is comfortably 60 MB, and almost none of that serves any purpose if the document is going to be read on a screen.
Compressing at 150 DPI typically removes 60–80% with nothing visible lost. The reasoning and the DPI table are in how to compress a PDF without losing quality.
One caution on ordering: compress after OCR, not before. OCR accuracy depends on resolution, so running it on an already-compressed scan gives measurably worse text for no benefit. Scan high, OCR, then compress the copy you send.
Frequently asked questions
What DPI should I scan at?
300 DPI for anything that will be OCRd or printed, 200 for text you only need to read on screen. Below 200 OCR accuracy falls sharply. Above 400 you are multiplying file size for detail nobody uses, unless the document is archival or has very fine print.
Should I scan in colour or greyscale?
Greyscale for black ink on white paper — roughly a third the file size with no visible difference. Colour only when there is colour that matters. Avoid pure black-and-white mode: it destroys text edges and turns faint pencil or stamps into solid black or nothing.
How do I scan multiple pages into one PDF?
A feeder does it by default. On a flatbed, look for "Add page" or "Scan more pages" in the dialog before saving — most software has it and most people miss it. If you already have separate files, merge them, watching the alphabetical sort order.
Can I fix a crooked scan?
Yes, but straightening re-renders every pixel at an angle, so text softens. A degree or two is invisible; ten degrees is noticeable. If the paper is still available, rescanning takes two minutes and costs nothing. Page rotation cannot help — it only does multiples of 90.
Is a phone scan good enough?
For text documents, yes — modern apps handle perspective correction, cropping and contrast automatically. The gap is illumination: a flatbed lights the page evenly, a phone captures whatever the room is doing. Indirect daylight, no flash, shoot straight down.
Should I compress before or after OCR?
After. OCR accuracy depends heavily on resolution, so running it on an already-compressed scan produces measurably worse text for no benefit. Scan high, OCR, then compress the copy you actually send.