>Why you don't use audiobooks ? It's an easy way to collect

>many hundreds of hours.

we do.

but the problem is that current acoustic model training scripts require segmented speech, and the process is not fully automated yet.