# Compress PDF Without Losing Text or OCR Quality

- Author: FastBankConvert (https://fastbankconvert.com)
- Published: 2026-10-17
- Category: Speed, size and limits
- Canonical: https://fastbankconvert.com/blog/compress-pdf-without-losing-text

> Compress a PDF without losing text: which settings shrink images safely, why some tools turn text into pictures, and how to check the result.
To compress a PDF without losing text, shrink only the images: lower their resolution to a level OCR can still read and keep fonts embedded. Avoid any mode that rasterizes, flattens or prints the pages to images, because that replaces real text with a picture of it. Then search the smaller file for a word to confirm the text survived.

## Can you compress a PDF without losing text?

Yes. In a digital PDF, text is stored as characters plus a font, and it takes very little space. Nearly all the size in a bloated file sits in images: logos, scanned pages and background graphics. A compressor that works on those images and leaves the text objects alone can cut the size sharply while every word stays selectable.

It helps to know what each part of a PDF costs. A page of statement text, stored as characters, is a few kilobytes. A full embedded font may add tens or hundreds of kilobytes, but only once per file. A colour scan of the same page can run to several megabytes. So when a statement is large, the question is almost always which images to shrink, and by how much.

The risk lies in the word "compress" itself. Different tools mean different things by it. Some resample images and leave text alone. Others redraw every page as a single picture at a low resolution, which is the fastest way to a small file and the surest way to lose your text. The rest of this guide is about telling those apart before you send the file anywhere.

## Image downsampling settings that keep scanned text legible

Downsampling lowers the number of pixels in an image. For a scanned statement, the page image is the only copy of the text, so it must stay sharp enough for OCR. Keeping scans at around 300 dots per inch in greyscale, or at a higher resolution in black and white, usually keeps characters readable while still saving space.

The table below lists the settings most compressors expose. The values are starting points to test on your own file, not fixed rules.

| Setting | Digital statement | Scanned statement that still needs OCR | Avoid |
| --- | --- | --- | --- |
| Colour image resolution | 150 dpi for logos is plenty | Convert to greyscale instead if colour adds nothing | Below 150 dpi on text pages |
| Greyscale resolution | 150 dpi | About 300 dpi | 72 dpi "screen" output |
| Black and white resolution | 300 dpi | 300 dpi or more | Dropping below 300 dpi |
| JPEG quality | Medium | High | Lowest quality, which blurs edges |
| Fonts | Keep embedded, subset | Not relevant to the page image | Converting text to outlines |
| Page output | Keep as PDF objects | Keep as PDF objects | Rasterize, flatten or print to image |

Two settings deserve a closer look. Converting a colour scan to greyscale often saves more than lowering resolution, and it costs OCR nothing on a black-and-white statement. Aggressive JPEG compression, on the other hand, creates soft blocks around each character that OCR can misread, especially digits like 3 and 8. The guide to [the best DPI for OCR](/blog/best-dpi-for-ocr) explains why 300 is the usual floor for statement print.

## Why some PDF compressors turn text into pictures

Some compressors reach a small file size by rasterizing every page: they draw the whole page, text and all, into one image and save that instead. The result looks the same at a glance, but the characters are gone. You cannot search or select the text, and a converter has to OCR a page that was perfect text before.

You will see this behaviour under several names:

- **Rasterize** or **convert pages to images**.
- **Flatten**, when it applies to the whole page rather than just form fields or annotations.
- **Print to PDF** through a virtual printer set to an image mode.
- **Maximum compression** or **smallest size** presets in some online tools, which do not always say what they are doing.

A rasterized statement hurts twice. First, the text is now a picture at whatever resolution the tool chose, often a low one. Second, the next step, converting to Excel, slows down because every page needs recognition, and the numbers it reads may not match the original. A file that was exact becomes approximate. If you need to get a statement into a spreadsheet quickly, [the fastest way to convert a PDF to Excel](/blog/fastest-way-to-convert-pdf-to-excel) starts from the original digital file, not a compressed copy.

There is one more variant to watch for: text converted to outlines. The page keeps vectors, so it stays sharp at any zoom, but each letter is now a shape. Search finds nothing, and conversion behaves as if the page were an image.

## Ghostscript presets for compressing a statement PDF

Ghostscript is a free command-line tool that can rewrite a PDF with smaller images while keeping its text as text. Its pdfwrite output device takes a preset called `-dPDFSETTINGS`, and the `/ebook` preset is a sensible first try for a statement: it reduces images to a medium resolution and keeps fonts.

A typical command looks like this:

```
gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -dNOPAUSE -dBATCH -sOutputFile=smaller.pdf statement.pdf
```

The [Ghostscript documentation for its PDF output devices](https://ghostscript.readthedocs.io/en/latest/VectorDevices.html) describes the presets and the image settings behind them. Its parameter table lists 72 dpi for colour and greyscale images under `/screen`, 150 dpi under `/ebook`, and 300 dpi under `/printer` and `/prepress`. The same page says pdfwrite goes to considerable lengths to keep the input's fonts as fonts in the output, which is exactly what keeps text selectable.

How to pick a preset:

- **Digital statement with a heavy logo or background:** `/ebook` usually shrinks it well, and the text is untouched.
- **Scanned statement you still need to convert:** start with `/printer`, which keeps images at 300 dpi. `/ebook` halves that and may hurt OCR on small print.
- **Never for a scan you need to read:** `/screen`, whose 72 dpi images are too coarse for reliable recognition.

The documentation also mentions a switch, `-dNoOutputFonts`, that turns text into linework or bitmaps. Leave it off. It is the command-line version of the "turn text into pictures" trap above. Ghostscript rewrites the file rather than editing it in place, so keep the original until you have checked the result.

## How do you check a compressed PDF for lost text?

Check the compressed copy before you delete the original. Search it for a word, select a line of text, zoom in on the smallest digits, and compare the page count. A file that passes all four still has its text, and a scan that looks sharp at high zoom will still read well under OCR.

Run these five checks on the smaller file:

1. **Search.** Press Ctrl+F (Cmd+F on a Mac) and search for a word on page one, such as the bank name. No match means the text layer is gone.
2. **Select.** Drag across a transaction line. You should get highlighted characters, not a single box around the page.
3. **Copy and paste a number.** Paste one amount into a text editor. It should arrive as the same digits you see.
4. **Zoom.** At 200 or 400 percent, look at the smallest print, such as the running balance column. Blurred digits or block patterns around them mean the image settings were too aggressive.
5. **Count pages and sizes.** Confirm the page count matches the original, then compare file sizes. A saving of only a few percent may not be worth any risk.

If the compressed file fails any of the first three checks, the compressor rasterized it. Go back to the original and choose a different tool or setting. If it fails only the zoom check, raise the image resolution one step and try again.

## When compressing a PDF before conversion is wasted effort

Compression is worth it only when the file is too big to upload or send. A digital statement of a megabyte or two gains nothing from compression, because the converter reads its text directly and the upload is already quick. Compressing it just adds a step and a small chance of damage.

Leave the file alone in these cases:

- **It is already small.** If the upload takes a second or two, the file size is not what is slowing you down.
- **The converter accepts it.** If it is under the upload limit and goes up without stalling, compression saves nothing you will notice.
- **It is the only copy of a scan.** Compress a copy, never the original, and only if you must.
- **You are about to download a fresh copy.** Getting the statement again from online banking gives a clean digital file that beats any compressed scan.

Splitting is often a better answer than squeezing. A year of scanned statements that is too big to upload as one file can be cut into monthly files, each of which goes up easily at full quality. You keep every pixel the recognizer needs, and a failed upload costs you one month instead of the whole year.

When an upload hangs, file size is only one possible cause. [Why a PDF upload gets stuck](/blog/pdf-upload-stuck) covers the connection and browser issues to rule out before you start shrinking files. And if the end goal is a file for another app rather than Excel, [converting a bank statement to CSV](/blog/bank-statement-to-csv) shows what that output should contain.

FastBankConvert is built to read a typical digital statement in seconds, so the bank's own uncompressed PDF is the best file to hand it. [The FastBankConvert homepage](/) shows the drop, convert and download steps in a sample run.