>>1044376
I have had that idea brewing for a while. llama.cpp just released 0.8.0 which includes a text-to-speech endpoint. Works easy enough in my own voice agent setup with LFM3. Just need to plug in an uncensored model and see how dirty it talks to me.