I've been experimenting all day with local language models and their NPC interpretation capabilities.

I have to admit I'm amazed at the ability of a local model to interpret a character and the creativity that a model like Mistral displays.

I have a lot of ideas, but it's terribly slow and misinterprets intentions.