Very nice. Sphinx is a great, but what we found was that the open source acoustic and language models are not good enough. To get any decent performance out of it you had to develop our own. We gave up at that point and switched our transcription process to a manual one. Maybe it time to go back and have another look at it.