Mozilla Common Voice releases the largest to-date public domain voice dataset, including 18 languages, almost 1,400 hours of data from 42,000 contributors.