Word-by-word captions in dozens of languages, burned into every format, plus an .srt file. Transcribed on your device; the speech model downloads once.