I'm not gonna pay for textshit anymore, and while I do regularly use local stuff that I can fit in my GPU the amount of rerolling, rewriting and fucking around I have to do in order to get around bad template phrasing has been getting old. Every time I ask for a model to try out that will fit in my 16gigs of VRAM none of it feels like it's genning anything that impresses me. I guess the problem is that I strictly want narratorial textgen, whereas most people I see that are pretty happy with it are just sexting with their waifu cards.

>>280290496
I was gonna save the art from that thread when it died normally, too.