I've just tried the demo. It's impressive, but not that impressive. Cloned voice sounds nothing like my voice and also there are subtle errors in the translation.
I'm not sure if it works much better than STT -> LLM -> TTS pipeline would. Also, this is of course just a first impression and not serious testing.
I'm not sure if it works much better than STT -> LLM -> TTS pipeline would. Also, this is of course just a first impression and not serious testing.