Skip to content
Account

Sample multi-page scanned PDF

This is a three-page scanned PDF with no text layer, for testing anything that has to work on more than one page: OCR, splitting, reordering, making a scan searchable. Each page is turned a different amount (0.7°, −1.1° and 1.8°) and the third is a little darker and noisier, as the last pages through a feeder often are.

3 pagesPDF 369 KBimage only, no text layer150 dpiCC0

Try it on this file

Read the text of all 3 pages

Opens the tool with this file already in it. Free, and it runs in your browser.

Or: Make it searchable (free) · Split it into pages (free)

Download

  • sample-scanned-report-3-pages.pdfPDF, 3 pages, image only · 369 KBDownload PDF

Free, no sign-up. Released under CC0 1.0: copy it, change it and use it anywhere. The exact values in it are in the answer key.

Page 1 of sample-scanned-report-3-pages.pdf, a fictional specimen marked SPECIMEN

Check your results

The exact values in this file, so your parser or OCR can be checked against them. They come from the script that drew it.

Table "2. Visits by quarter", page 15 rows of Quarter · Days open · Visits · Per day; the first is Apr - Jun 2025 · 74 · 9,620 · 130
Table "5. Money", page 38 rows of Line · Income · Spending; the first is Council grant · 62,000 · -
Text61 lines on 3 pages, the first "Larkfield Community Library"

Answer key, JSON · 6.9 KB

What filedone reads from it

The first lines Scanned PDF to text returned for sample-scanned-report-3-pages.pdf, as they came out. 535 of the answer key's 554 words came back exactly.

--- Page 1 ---
Larkfield Community Library
Annual report, April 2025 to March 2026
A fictional libr ary, written for testing OCR. No real place or person is beer ibed.
1. The year in brief
The library was open on 302 days this year, six more than last year, after Saturday
afternoons were restored in September. Visits rose to 41,280, an average of 137 a day.
The busiest month was December, when the winter reading club ran roa evenings a

Recorded on 31 October 2026 with the tool's default settings, on its server version (Tesseract 5.5.3, English model). Your browser runs the same engine with the same settings, so your result can differ slightly. Try it on this file

What is in the file

The annual report of a fictional community library: headings, paragraphs of running text, a table of visits by quarter on page 1 and a table of income and spending on page 3. Income and spending both total 73,570, so a converted table can be checked by adding it up.

The pages are 150 dpi, a common setting for a document feeder, which makes the letters smaller in pixels than on the 200-dpi single page. It is a harder read on purpose.

The facts

Measured from the file itself by the script that made it.

Pages3
Filessample-scanned-report-3-pages.pdf (369 KB)
Text layerNone: each page is an image only, so a tool needs OCR to read it
Page sizeA4 (210 × 297 mm)
Resolution150 dpi
ColourGrayscale
Skew by pagepage 1: 0.7° anticlockwise; page 2: 1.1° clockwise; page 3: 1.8° anticlockwise
LanguageEnglish
FictionalEverything: every company, person, address, number and figure is invented, and every page says SPECIMEN
LicenceCC0 1.0 (public domain)

What to test with it

Licence, and who is in it

Nobody. Every company, person, address, account number and figure in the file was made up by the script that drew it, and every page carries the word SPECIMEN and a line saying it is fictional; the PDF's metadata says so too. The file is dedicated to the public domain under CC0 1.0, so you can use it in tests, demos, courses and products without asking. Last updated 22 October 2026.

Questions

Do the numbers in the tables add up?

Yes. On page 3, income and spending are both 73,570, and on page 1 the four quarters' visits add up to the year's 41,280. That lets you check an extracted table against itself.

Why are the pages skewed in different directions?

A tool that straightens a whole document by one angle gets two of the three pages wrong. Different angles per page test whether each page is handled on its own.

Can I use this file in my own tests, product or course?

Yes. It is released under CC0 1.0, a public-domain dedication: copy, change, redistribute and use it commercially, with no need to ask and no need to credit filedone (a link is welcome).

Is anything in it real?

No. Every company, person, address, account number, tax number and figure was invented by the script that made the file. Every page says SPECIMEN and that it is fictional, and so does the PDF's metadata.

Open it in

Guides

More sample files