Calibre is free, local, and it can do things no web converter does: edit the EPUB afterwards in Sigil, manage a large library, convert on a schedule, output KEPUB. If you already have it and you like owning your tools, keep using it.
Installing it, learning the conversion settings, and accepting the output quality. Calibre's PDF
input is the weakest part: it does not have a document structure model the way Word or InDesign
does, so it has to guess. Two settings do most of the work: --enable-heuristics (which
removes headers and footers, unwraps lines and removes hyphens) and --output-profile
kindle for device-sized output. Even then, the paragraphs sometimes come out wrong and you
end up fixing the file by hand in Sigil.
One file, no install, no library to maintain, and nothing left on your machine. It is the right tool when you have one PDF of a textbook and you want it on the Kindle tonight. It is the wrong tool when you have four hundred of them and you want a pipeline.
Calibre needs an OCR plugin wired up and configured before it will read a scan, and the result still needs cleanup. OCR built into a converter is one click. If your book is a scan, that is the difference between an evening and a weekend.