Quoted By:
After messing around with locals/colabs the entire day and trying out XWin-Mlewd, Mythomax and CC-Mistral with a bunch of different settings, I think I'm going back to NovelAI. It has issues and it's still a 7b model, but I think the output is generally more high-quality, it comes bundled in with a bunch of useful presets, covers a wide array of use cases comes with the added bonus of lightning-fast response times. But honestly, if you have 30$ a month to spend on this, you might as well go all-in, add a little bit of money to the pile, and rent out a powerful VM for some super-kino 70b model.