If you have a good GPU there's open-source models like Mixtral you can use locally on your computer for free, which are as good as ChatGPT 3.5 while also being uncensored. Considering filian's 3D model isn't even custom, you could easily put it into something like this and use another AI to clone her voice.
https://www.youtube.com/watch?v=6HSWMHzQ7TEGo to /lmg/ on /g/ if you want to learn more.