i had tried chardet with dataset on non-US files.
https://www.kaggle.com/datasnaek/youtube-new/downloads/youtube-new.zip/40

its detects windows-1254 instead of correct utf-8 as reported by linux file utility