This is fantastic. It would be great to use this to also train text-to-speech models replicating specific people by feeding unsubbed video clips of the same pe

This is fantastic. It would be great to use this to also train text-to-speech models replicating specific people by feeding unsubbed video clips of the same person into it, a la something like wave net [1], or what Google is doing with John Legend [2]. [1] https://deepmind.com/blog/wavenet-generative-model-raw-audio/ https://deepmind.com/blog/wavenet-generative-model-raw-audio... [2] https://www.youtube.com/watch?v=AgHghcYqeto https://www.youtube.com/watch?v=AgHghcYqeto