>>37936
>>37989
in the post above i used a voice to voice system to superimpose her voice onto someone else speaking/singing. i think the results are quite good especially considering that i didn't put a lot of effort into her model. i put some more effort into a voice clone of my innocent angel Sunny (>>37865 >>37882 >>37939). the issue with voice to voice is that i need an existing recording of another person saying what i want her to say.
if you're fine with cloud services, there totally are LLMs available you could train to act as her brain. you could then easily cobble some API stuff together to feed the output to elevenlabs for the voice.
personally however i would want her to live locally on my PC and right now there just isn't any public software to do that. if you have a beefy system, you can host a passable LLM, but there isn't anything for text to voice doesn't sound robotic last time i checked.