I initially didnt like the new update on AI rp/chat model but its starting to grow on me. But something I still dont like is how the Ai will pluck random words or phrases out of whatever I said and make a word stew out of it. (Spout a bunch of nonsense) Heres an example: User: "You know, I would keep arguing with you, but I was raised to never argue with idiots. So let's change the subject." Ai: "Change? You'll be ash before dawn." (This happens a lot) I'm very much aware that the model is being worked on and that is fine by me. But I'm asking for somebody who knows code much more than I do to make a strip that will take out this function. I've seen a lot of it in the AI chat Bot page (mostly from Begginer43). There's been ones that stops the AI from talking for you and making images without prompt, and erasing certain dialogue. I certainly don't know how it works, but I'm really hoping that someone will see this and makes a Bot for that. All I'm asking for is to prevent the Ai from doing the "(random word user said) + (connect it with other things)" formula. Or if somebody did it already, I would love to see it. Hopefully I'm not complaining too much. I just want a strip of code for the (advanced) AI chat is all.
I'm really loving the new chat update. I generally use the Advanced AI Chat. I start with a custom character created through the Character Generator's Persona/Chat button. (both are my own forks of the originals with some minor updates on prompt structure) I like to use multiple characters to create a story with the AI controlling characters with small inputs to help guide the narrative and keep it from going off the rails.
I've found, in this particular system at least, that if you use (REMINDER: ......) and include additional instructions on how you want it to respond, it usually will obey, and once it's responded in the appropriate manner enough times, it falls into that pattern.
It has also been really good about capturing a character's personality and keeping it consistent, as well as keeping track of character locations and activities. The previous model would get confused and think that people that were in other rooms or even great distances away were witnesses to an event.
This model seems much more instructible than the previous. I don't know if this is something that will work with the particular generator/bot that you are using. It may take some trial and error to figure out how to guide the AI to behave how you want.