One PDF Into A TXT File
The converter first checks whether your PDF already carries a text layer, since many PDFs exported from a word processor or a website do, and if so extracts that text directly rather than running OCR on it at all. If the PDF is a scan, each page is treated as an image and OCR reads it before the words are written to the output file.
Either way you get one .txt file with the pages appended in order and a separator marking each page break, ready to save, search or feed into a script; the distinction between extracted and OCR'd text is invisible in the final file.
Why Check For A Text Layer First
Running OCR on a PDF that already has embedded text would waste time and risk introducing recognition errors into content that was already perfectly accurate, so this tool checks first and only falls back to OCR for pages that are genuinely scanned images with no text underneath them.
That matters for mixed PDFs too, where a cover page might be a native document and the following pages are scanned attachments: each page is handled the way its own content calls for, rather than the whole file being forced through OCR uniformly.
How To Tell What You Have
If you can already select and copy text from the PDF in a normal reader, it has a text layer and the .txt output will match that embedded text closely, since no OCR guessing is involved at all; accuracy in that case depends only on how the PDF was originally created.
If clicking to select text just highlights a rectangle around the whole page, or nothing happens, the page is a scanned image and the .txt content for it depends on OCR quality; check scan resolution and the sharpness of the original document if a scanned page reads worse than expected. A table inside a PDF that already carries a text layer usually keeps its cell values in a sensible reading order, since the position of each character is stored directly in the file, while the same table on a scanned page depends on OCR walking the columns correctly, so a scanned invoice is more likely to need manual realignment afterward than one pulled straight from a native PDF.
Getting Reliable TXT Output
For a scanned PDF, a higher-resolution scan going in produces cleaner text coming out, with 300 DPI or better a reasonable target if you are the one doing the scanning, since anything lower starts to cost OCR accuracy on smaller print.
For a PDF that mixes native and scanned pages, check the .txt output around the boundary between the two, since that is the one place page-level handling changes and where an unusual header or footer sometimes confuses the page-order in the separator markers. The file is written as UTF-8 with Unix-style line breaks throughout, whether a given page came from direct extraction or from OCR, so a script reading the result does not need separate handling for the two kinds of source page once the text has been written out.