Check CMU Sphinx. Its quite dated though. To get any decent performance out of it you'll have to train it and generate the databases (Acoustic Model and Languag

Check CMU Sphinx. Its quite dated though. To get any decent performance out of it you'll have to train it and generate the databases (Acoustic Model and Language Model) yourself. Google has not open sourced their Engine or databases. I don't think it will happen. There are no ready made solutions for auto-generated transcriptions. Speaker Independent Speech Recognition is still one of those unsolved problems and unlikely to be solved anytime in the near future. Even research is at a standstill. Check http://news.ycombinator.com/item?id=1313679 http://news.ycombinator.com/item?id=1313679 for more.