Models in this directory are exported using --decode-chunk-len 96
instead of --decode-chunk-len 32.
The advantage is that it runs much faster, while the disadvantage is
that you will experience some delay before you see the recognition result
after you say something.