Although the Gilder/Levin method works like a charm and hence looks like a viable image replacement option, some other issues remain (besides the non-semantic span tag). The most important ones not being able to select, copy and resize the text and secondly the inability to easily alter the text, which is now enclosed in an image.