Conversation

Notices

1. LS (lain@lain.com)'s status on Tuesday, 06-Feb-2024 17:28:49 JST LS

   • kaia
   @kaia elevenlabs voices made form actual speech samples are usually better at capturing the speech style. for base models, you often have to crank up the 'expressiveness' slider and that can become very unnatural too, so you have to regenerate until you get a good take.
   • LS (lain@lain.com)'s status on Tuesday, 06-Feb-2024 17:29:46 JST LS

     in reply to

     • kaia
     @kaia and there's also the possibility of just changing the voice and not wholly generating it, that will keep the basic speech patterns of the original speaker