Comment by gsjbjt

That's awesome - can you say anything about what datasets this was trained on? I assume something specifically conversational?