PDF to text, from a scan or from a PDF's own words
The words, out of the file
Drop a PDF on the OCR page, choose to make only the text, and read it. A page whose words are already text is quoted as it stands, and the result says so: those words were not read here, and when another program put them there, they were not checked either. A page that is a picture of text, a scan or a photograph, is drawn at 300 dots per inch and read. Download the text three ways: as it reads, a page after another and a column after another; laid out as the page has it, each line where it stands; or, for a page that is clearly a table, as a CSV for a spreadsheet, whose cells can never run a formula. A page that is not laid out as a table gives no CSV, and the result says so rather than handing you a spreadsheet of misplaced cells.
Sure, or said
A wrong word under a scan is worse than none: a search finds the wrong place, and a copy carries a mistake nobody typed. So a word read with too little confidence is left out rather than guessed, and the result says how many were left out on each page. A page on which nothing can be read with confidence is given no words at all, and is named as unreadable. When no page can be read, nothing is made, and you are told why.
Languages
Choose the language, or let the page find it: it looks at the writing first — Latin letters, Cyrillic, Greek or another script — and then tries the languages written in it. Albanian, English and the other languages of the region are among them, and an accented letter read comes out as that letter, not as a dot or a question mark. How well a language is read was measured on invented pages for a few of them; every other is read, not measured.
A searchable PDF as well
The same reading can put the words under the picture of each scanned page, so the PDF itself can be searched and copied from, with the picture left exactly as it was. OCR PDF says how.
Get the text from a PDF now — it opens the tool, with nothing to install.
The record
Reading changes nothing in the file you drop. A searchable copy is a new file, and its record says that a text layer was added, in which language, what became of each page and how many words were left out, and that the original picture is preserved. The text is kept with the file it was read from, and deleted with it or whenever you delete the reading; your account's trail says what was read and how many words, never the words. Files of up to 50 pages are read, at the size your plan uploads, and deleted after 90 days, or whenever you erase them. What a download costs is on the pricing page.