I use vim. Minimum of keypresses, maximum effect, plus regexp is a pretty integral part of the editor.
Besides, I'm convinced this GUI-WYSIWYG-thing is just a passing fad
I also use sed extensively, primarily because of its multi-line capabilities (e.g. commenting out first line on each page when it's a page header, joining words split over lines etc.), and sometimes awk.
Have accumulated a rather nice set of bash scripts for assisting the whole process of converting a book from scans to epub.
(Well, I'm deviating a little from the truth here. I also generate an html-document with the scans and ocr-text side-by-side, which I import into libreoffice writer for proofreading. Or convert to a PDF for proofreading on the tablet.)