Turning paper into a usable PDF with nothing but a phone

What a scan is, and what a photograph is not

Four differences separate the two, and every one of them is fixable with a phone if you know which you are fighting.

• Perspective. A scanner looks at the page straight on. A phone held at an angle sees a trapezoid, with the far edge compressed. Auto-crop can find the edges, but it cannot undo the foreshortening perfectly.
• Lighting. A scanner applies the same illumination to every square millimetre. A phone inherits whatever the room does, which is usually one bright source above and behind you, casting the shadow of your own hand across the page.
• Contrast. Scan software thresholds the page — pale paper becomes white, ink becomes black. A phone records what is there, so paper comes out grey and the ink-to-paper contrast is far lower than the eye perceives.
• Framing. A scan is exactly the page and nothing else. A photograph includes the desk, the edge of the notebook next to it, and your thumb.

Light is the whole problem

If you fix one thing, fix the light. Nearly every bad phone scan is a lighting problem that looks like a camera problem.

• Do not use the flash. It is a point source a few millimetres from the lens, so it produces a bright hot spot in the middle of the page and a falloff toward the edges. It also reflects off any glossy paper into a white glare band.
• Do not stand under the ceiling light facing the page. You will cast your own shadow directly over the part you are trying to capture, and it will land wherever your hand happens to be.
• Use two diffuse sources at roughly 45 degrees — daylight from a window plus a second lamp works well. Two sources from different directions cancel out each other's shadows, which is the entire trick.
• Indirect daylight is the best free option. Near a window but out of direct sun, the light arrives from a large area rather than a point, which is exactly what a scanner uses.
• Glossy paper needs an angle change, not an exposure change. If you can see a reflection of a light, tilt the phone or the paper a few degrees until the reflection moves off the text. Brightening the image will not remove glare.

Square to the page, or fix it later

Hold the phone directly over the page, with the lens axis as close to perpendicular as you can manage, and keep the page flat. Every degree of tilt compresses one end of the text and makes it smaller than the other end, and no crop can undo it.

A stack of sheets can be pressed flat under a book at the edge, or held with a clip. A bound book is harder: the closer a page is to the spine, the more it curves, and curvature changes the effective scale across the line. Shoot bound pages as flat as the binding allows, and accept that the last centimetre near the spine will always be the weakest part of the image.

Framing matters more than people expect. Fill the frame with the page, but leave a small margin of desk visible on all four sides. Automatic edge detection needs to see where the paper ends in order to crop it; a page that runs past all four edges of the photograph gives it nothing to find.

Camera settings that help, and the ones that quietly hurt

• Turn off digital zoom. It crops the sensor, so zooming in throws away resolution and adds nothing. Move the phone closer instead.
• Use the main camera, not the ultra-wide. Ultra-wide lenses bend straight lines near the edges of the frame, which is exactly where the text on a full page sits.
• Turn off portrait and beauty modes. They identify edges in order to blur what is not the subject, and on a document the edges are the content.
• Consider turning off HDR. It is designed for scenes with a bright part and a dark part; on a flat, evenly lit sheet it tends to lift the paper toward grey and reduce the contrast between ink and paper.
• Shoot at full resolution. You can always make a file smaller later; you cannot recover detail that was never recorded.
• Capture in colour even if you want greyscale. Colour to greyscale is a one-way conversion you can perform any time; the reverse is not possible.
• Tap to focus on the text rather than letting the camera choose, and check that the result is sharp at 100 per cent before you move on to the next sheet.

Shooting a stack without losing the order

A scan job is usually more than one page, and the mistakes that ruin it happen before any software is involved.

• Shoot one page per photograph. Two pages in one frame halves the effective resolution and makes cropping ambiguous.
• Work through the stack in one session, in one place, with the same light. Mixing a morning batch and an afternoon batch produces a document with visible bands of different colour and exposure.
• Check each shot before turning the page. Retaking one page costs seconds; discovering a missing page after the paper has been filed costs a trip.
• If a page is landscape, either rotate the page itself or note it — do not rely on remembering which one it was.
• Count the sheets before you start and count the shots when you finish. A missing page is the single most common defect in a phone-scanned document, and the only one nobody notices until it matters.

From photographs to one PDF, in the right order

Doing these steps out of order creates work. Cropping after you have resized a page, for instance, means the crop and the scale no longer agree.

• 1. Crop and straighten each page. Remove the desk, the thumb and the shadow. This is the step that turns a photograph into a document, and doing it first makes every later step easier.
• 2. Rotate the pages that came out sideways. Fix orientation now rather than after assembling the file, so you are never rotating a page you have to find again.
• 3. Put the pages in order. If the shots are named in sequence, numbering them 001, 002, 003 rather than 1, 2, 3 keeps the order correct past the ninth page — and it is the order the merge tool will follow.
• 4. Combine them into a single PDF. One file per document, in reading order.
• 5. Check the result. Open the finished file and page through it once. Count the pages against the sheets you counted earlier, and look at the one page nearest the spine or the one that was hardest to shoot.
• 6. Compress only if it needs it. Phone photographs are large, so a twenty-page stack can easily reach 60 MB. If the destination is email, compress a copy and check that the smallest text is still legible before you send it.

How much resolution you actually need

A modern phone easily exceeds 300 dpi when a page fills the frame — a 12-megapixel main camera gives roughly 3,000 by 4,000 pixels, which over an A4 page works out around 350 dpi. The resolution is rarely the limiting factor. Light, angle and flatness are, which is why the same phone produces excellent and terrible scans on consecutive attempts.

The practical consequence is that you almost always want to reduce, not increase. A stack of full-resolution phone photographs is far larger than it needs to be, and downsampling to 150–200 dpi for anything headed to a screen is the single biggest saving available.

If the text has to be searchable

A PDF built from photographs contains no text, only pictures of text. It cannot be searched, and nothing in it can be copied. That is a property of the method, not a defect in the file.

Producing a searchable version means running OCR over the pages, which needs a real engine rather than a browser tab — the models are far too large to load reliably on a phone. Most scanner apps bundled with phone makers do offer it, and the built-in scanning features of recent phones increasingly do as well. If searchability matters, capture the pages as well as you can and run OCR through whichever tool you have access to, rather than trying to fix it afterwards from a low-quality image.

A clean, evenly lit, straight shot is the entire input to a good OCR result. There is no setting inside an OCR tool that repairs a photograph taken under a ceiling light with the phone tilted.