Thanks a lot! This should be fixed now.

I started experimenting with the speech recognition browser API yesterday (so that the user can listen and repeat sentences), but it's not supported everywhere.