>>101003145NTA, while I doubt that Evil could choose herself currently (that is to say, the voice output vedal is using just decided that that stream would be the stream to do it more often) but I can see Vedal testing out trying voice tokens aligned for Evil's LLM output to then feed into the voice output.
But then again, we would most likely be seeing similar crap we used to see in 2023 with the ssml or <|token|> stuff
