I just wrote a character-based diff that handles UTF8 and also displays binary (non-UTF8) bytes in-line. Is this something you would use?