SMI is HTML-like and many players outside Windows and Korea do not read it. Plain text opens on every device and is easy to search or paste elsewhere. SMI is at home in subtitles for Korean-market players and multilingual files; TXT is the usual choice for notes, logs and text that must be read by anything.
The text is read as UTF-8 or UTF-16, whichever the file is. An older file in another code page is read using the charset it declares, or as Windows-1252 if it declares none, so accented letters in an unlabelled legacy file can come out wrong. A file that is not text, such as the picture-based subtitles of a DVD, is refused instead of being converted to nonsense. A SAMI file can hold several languages; only the first one found is read. The result is a transcript with one cue per line, without times; a cue's own line breaks become spaces.