Nice and really needed project! What's the average latency overhead when translating, in seconds? And, I believe, some language phases need to be pronounced in full in order to be properly translated, so I'm also wondering how your realtime translator handles that – does it mean dynamic latency? If so, what's the minimal latency?