Curious if anyone has practical real world input on training CMU based ASR engines (Sphinx, PocketSphinx) and / or creating and tuning voices for the TTS related components. Just trying to understand how hard it is, what the realistic gap is to use these tools in real world applications.