I am working on a plugin for KaDoSu to index audio files. I found these three free libraries Sphinx, ISIP and Julius for speech recognition and decided to use Sphinx because it is a pure Java implementation and so I don’t depend on natively compiled libraries. But all these libraries have the same problem. They need to be trained to recognize speech well. Voxforge collects speech samples from its users and shares them for free with all these projects. Please donate your voice to get better free speech recognition.