Why Is My Scanned PDF So Large?
Your 5-page scan is 14 MB and a 200-page ebook is 2 MB. Here's the pixel arithmetic behind that, which settings cause it, and what to change next time.
A scanned PDF is large because it does not contain a document — it contains a photograph of one, page by page. An A4 page scanned at 300 DPI in colour is 2,480 × 3,508 pixels, which is 8.7 megapixels at 24 bits per pixel, so that single page is about 26 MB of raw data before compression; the scanner squeezes it down with JPEG, but 1–3 MB per page is a normal result, and five pages of that will beat a 200-page ebook.
If the file just has to get under an upload cap, re-encode those page images at lower resolution and colour depth. Convertlix for Android does that on the handset with its Compress PDF tool, which shrinks PDFs by re-encoding images; files are opened through Android's own file picker and never leave the device — no upload, no server copy, nothing to retain. The scope: Android 8.0 or later, sign-in required (no guest mode), files up to 25 MB per job on the free tier against a server-set monthly allowance of conversions shown in the app, output in Downloads. The rest is why it got so big.
You're not imagining this: the file dwarfs the document
A five-page scan at 14 MB looks absurd next to a 40-page report at 300 KB. Nothing went wrong: the two files store different things, and only one is storing your document.
- You can change how the pages were captured — resolution, colour mode, encoding, cropping. That is nearly all of the size.
- You cannot change that a scan has no text in it. No setting turns a photograph of a page back into characters and fonts; OCR reads the letters and writes them alongside the image. That floor is why a scanned page never reaches the size of a born-digital one.
The arithmetic, page by page
Every scan decision is a multiplication, which is why the numbers escalate.
Resolution squares. DPI is per inch in each direction, and an A4 sheet is 8.27 × 11.69 inches. At 150 DPI that is 1,240 × 1,754 pixels — 2.2 megapixels; at 300 DPI, 2,480 × 3,508 — 8.7 megapixels, four times as many; at 600 DPI, 4,960 × 7,016 — 34.8 megapixels, sixteen times the 150 DPI page.
Colour depth multiplies again. Full colour is 24 bits per pixel (three bytes: red, green, blue), greyscale is 8, and bilevel — pure black and white — is 1. So a 300 DPI page is about 26 MB raw in colour, 8.7 MB in greyscale and 1.1 MB in bilevel.
Encoding decides what survives. JPEG is built for photographs: smooth gradients compress well, sharp black-on-white letter edges do not, and it leaves faint halos around text. CCITT Group 4 and JBIG2 are built for scanned text and reach 30–80 KB per page, because print is mostly identical white pixels.
A text PDF stores something else. A born-digital export stores the letter "a" as one character plus a font reference — a few bytes — which is why 200 pages of text sit under 2 MB while five scanned pages sit at 14 MB.
Scan setting to size per page
Approximate ranges for a normal typed A4 page; your scanner's JPEG quality setting can move any row by a factor of two.
| Scan setting | Typical size per A4 page | What you lose |
|---|---|---|
| 300 DPI colour, JPEG | 1–3 MB | Nothing |
| 300 DPI greyscale | 400 KB–1 MB | Colour: stamps, ink, highlighter, seals |
| 200 DPI greyscale | 150–400 KB | Detail below about 8 pt type softens |
| 150 DPI greyscale | 80–250 KB | Print sharpness; still clean on any screen |
| 300 DPI bilevel (CCITT G4 / JBIG2) | 30–80 KB | All shading; photos and grey stamps unusable |
Most oversized scans come from two choices: "photo" mode left on, and colour left on for black ink on white paper.
Image-only versus text, and what OCR does
An image-only PDF holds one picture per page; a text PDF holds characters. A searchable scan is both — the picture, plus an invisible text layer that OCR produced by reading it.
That layer is only a few kilobytes per page, so OCR makes the file marginally larger, never smaller. It is worth running because it makes the scan searchable and selectable, but the bytes have to come from resolution, colour depth or encoding.
How to scan smaller next time
- Choose the DPI first. 300 DPI is what many portals ask for and the practical floor for accurate OCR of small print. 200 DPI is comfortable for typed documents read on a screen, 150 DPI fine for a receipt. Go to 600 DPI only for fine artwork.
- Switch to greyscale unless the colour carries information. Contracts, transcripts and utility bills lose nothing and shed 40–60% of their size. Keep colour for passports, ID cards, cheques and official stamps.
- Turn off "photo" or "best quality" mode. Those presets assume a photograph: they keep gradient detail your page does not have and skip the encoders built for text. "Document" mode also flattens off-white paper to clean white.
- Crop to the page edges. A4 on an A3 flatbed stores the empty half of the glass, and a phone photo stores the desk and the shadow. Crop before you build the PDF.
- Use the camera only as a fallback. With no scanner around, a tool that scans pages with your camera — Cam to PDF, in the Android app named at the top — works: lay the page flat, fill the frame, and avoid one hard light source. Turn Photos Into One PDF for an Upload (Android) covers assembly.
- Only then compress, and re-check the number. If the scan is still over a cap, re-encoding the page images is the fix: Compress a PDF Under 2 MB on Android (Free) covers the order of operations, Compress a PDF Under 500 KB Without Blurry Text the tighter caps.
When a large scan is the right call
Not every oversized scan is a mistake. Leave it large when:
- It is an identity document. A passport page, ID card or visa sticker is checked for colour, holograms and fine security printing, and a reviewer who cannot verify a stamp rejects the application.
- A coloured stamp, seal or signature is the point. Notarised documents, bank letters and transcripts are often refused on that detail alone.
- You are scanning to archive. If this is the last time the paper exists, scan at 300–600 DPI in colour and keep that master; smaller copies come from it, never the reverse.
- It contains photographs, artwork or medical images. Bilevel and heavy JPEG destroy continuous tone, and nothing later brings it back.
Scan once at the quality the document needs; compress copies for whatever the upload form needs.
FAQ
Is 300 DPI too much for a scanned document?
For a document you only need to read on a screen, yes — 200 DPI is usually indistinguishable and roughly half the size, and 150 DPI is fine for receipts and simple forms. 300 DPI is right when a portal specifies it, when the print is small enough that OCR accuracy matters, or when the page will be printed. Above that you pay in size for detail almost nobody looks at.
Does OCR reduce file size?
No — it makes the file slightly larger. OCR reads the letters in the page image and writes an invisible text layer alongside it, a few kilobytes per page, while the original image stays where it was. It is worth running because it makes the scan searchable, but size reduction has to come from lower resolution, greyscale or re-encoding the images.
Why is my 5-page scan bigger than a 200-page ebook?
Because the ebook stores text and the scan stores photographs of text. The ebook records each letter as one character plus a reference to a font shared across the book — a few bytes. The scan records that same letter as several hundred pixels, and one A4 page at 300 DPI in colour is 8.7 million pixels, about 26 MB before compression.
Does Convertlix upload my file?
No. Files are opened through Android's own file picker, processed on the phone and never sent anywhere; the app asks for no storage permission, and its one runtime permission is the camera, used only for scanning pages. Its network calls carry identity, licence, quota, configuration and payment data, never a document — sign in once, turn on airplane mode, and the tools still work.
Does scanning in black and white ruin the document?
For black ink on white paper it usually looks better than you would expect, and the saving is large: bilevel encoding with CCITT Group 4 or JBIG2 often reaches 30–80 KB per page with crisp letter edges, against 1–3 MB for the same page in 300 DPI colour. It destroys anything with shading — photographs, grey logos, faint pencil, coloured stamps — so check every page first. A desktop scanner's own driver usually exposes bilevel mode directly, and Ghostscript gives the same control from a terminal.
If the scan you have is too big for the form in front of you, re-encode a copy locally, keep the original as your master, and check the size before you upload.