currently included german and english. contributions for other languages welcome, native speakers will have better insights into quality of speech output and recognition model