The new design looks very good. I could definitely see this being useful for online educational videos and interviews. I often find that I can read faster than the conversation happens on video. It would be very cool if they could partner with Khan Academy to transcribe all of the lessons. This would allow for translation to other languages, which would also be a cool feature to build into the product.
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.
Helps to have custom models. The free ones though are decent and a lot of recent sphinx improvements help.
I preferred speakertext when it cost $.30/minute, not the new price of $2/minute. I can understand why they would want to increase the price, but the new price point means that it is not usable by casual bloggers.