[nphrk]

Very nice idea! I have one technical question - do you also use some speech-to-text software, or is it 100% human labor? I believe the utilization of such software (even if it gives crude results) can be of great use to a service like yours.

[braindead_in]

Right now it's all done manually. I experimented with CMU Sphinx II when I started but results were poor. The main problem was actually the recording quality, eg. background noises, people talking over each other etc. I plan to revisit it once again and try. Maybe things have improved.