I\'ve got a text to speech model set up that takes Polish speech, changes it to Polish text then converts it to English text then speaks it, but it doesn\'t work the first t