I'm currently using gryphe/mythomax-l2-13b:free
with the only rule being: "Make sure your responses are only a single sentence. Don't repeat responses you have already used."
It still breaks the single sentence rule but it makes sense whenever it does, such as giving an explanation or a story.

I am currently having a blast with just these 2 rules on this model. The AI seems to have extremely varied conversations and responses with it.

For example somebody on a server was talking to a medic, the medic responded:
"We gotta get a welder and patch them wounds up!"
and then both of them got into an argument about using a welder for medical purposes with the AI responding: "Desperate times cause for desperate measures! Now its time to use a welder for those wounds!"

I also got told in a station: "Sir, there are no clowns here. Now leave before I take you out with the garbage.

And a mechanic stole my chair calling himself the "legendary repairs man" and told me that he's earned that chair.


Even though gryphe/mythomist-7b:free seems like a strait upgrade, I had a lot of problems with the AI repeating itself and never going past a single topic.

Example: "I don't know (insert literally any topic here) captain. But we must work together and have great teamwork to survive!"



Anyway, I was wondering what settings/models anybody else has used for hilarious results.
There is sort of a sense of realism when half of the AI is professional and rude while the other half is mentally unstable.