Small update about llm/ai text smut:
While googlegemini is still worse than chatgptlatest and especially claudeopus/sonnet3.5 it's completely free (unlike them and other big corpo models) and the latest update (GeminiExp1114) finally made it worth using. Also gemini2 (which should be even better as this model is probably just a preview version of it) should be released in december.
All you need is to register a google account, use vpn (or be in one of the countries they allow free usage via api) grab api, install ST (or risuai, agnai, etc.), enter your api in makersuit window. It has a filter but it's not internal (internal is practically nonexistent), i.e it's a moderation endpoint that acts very schizo, kinda insane, e.g it can be hard to break it for non english languages or the very first messages (especially when using system role) or... if you use a small amount of long messages instead of many shorter ones that have the exactly same context as the longer ones (yes it's really schizo), but once you started, it's very easy to not trigger it later. You can also disable streaming for better results.
Also Musk's Grok is free untill the end of the year, but it's probably even dumber, still should be superior to all local slop models, except (maybe?) latest nvidia's nemotron finetune and some other model I forgot about.