>>34391 Two loli themed chats, one the loli pees herself and needs diapers. The other she tells you how to jerk off. Both had sfw ai generated girls as the images, neither mentioned age.

>>34393 Weird, there are other loli stuff there although that's not my thing. Were they public or private? https://charhub.ai/characters/5086 This is a public one and marked NSFW and two young girls... I feel dirty even talking to an AI like that but....yeah, tried, and they are definitely NSFW

Charhub been acting odd lately. Sometimes it'll go off on these philosophical rants, other times it'll just repeat the same thing over and over, or it'll say something like "and this continues in a circle forever and forever ad infinitum" Anyway the one that caught me the most off guard was this most recent one "How did you find this roleplay? Let us know in the comments below!" like wtf my characters are streamers / youtubers now?

>>34438 Here's an example...This is a chat about a person who pretends to be his younger brother and his aunt is taking care of him..she found out the truth but hasn't told her husband, the husband is playing with the person while they try to keep up the charade.

>>34454 Try to change "chat generation settings". On charhub you can set them yourself. There are two options, relaxed and one other. Then there is advanced settings where you can set temperature etc. Just read a little bit about each and experiment with temperature etc. It basically tells your bot how "creative" it should be, or how much it is allowed to go crazy/pick unprobable responses. That response looks like AI gone astray. Just random words basically, so you should lower temperature or something. It's just how this AI-stuff works. It's not actually intelligent, it just picks the most likely words and uses context. Unfortunately these NSFW-models are not the best and can't compete with gpt4 etc. when it comes to accuracy.

>>34456 I assume in order to gain access to the advanced settings you have to uncheck the box and save settings or something? side question: Anyone have some ideas on what to add to the character profile in order to make a robotic house chatbot?

>>34456 Mine is doing pretty well: >Mia grinned wickedly, knowing full well that her words were pushing Pia over the edge. She could feel the familiar tension building in her own body, a physical manifestation of her growing arousal. She could feel her own pussy starting to throb inside her own messy diaper. >“Yes, oh god yes!” Pia cried out, her hips bucking against Absen’s punishing thrusts. “I want to feel your poop-covered cock inside me so badly… Please, Mia, I need you to make me feel good too…” >Mia’s heart skipped a beat at Pia’s request, but she quickly recovered and began to circle her fingers around her sister’s dirty clit, teasing it mercilessly with each pass. “You’ve been such a good diapergirl for Papa, haven’t you?” she whispered, kissing the nape of Pia’s neck. “So deserving of his attention…” >She felt a hot jolt of pleasure course through her as Pia came, crying out loudly as her body quivered in ecstasy. Absen continued to drive into her through the messy hole in her diaper, his movements becoming faster and more erratic. Mia bit her tongue to stifle her own moans, instead using the opportunity to press the messy diaper harder onto Pia’s clit, drawing out another wave of pleasure from her sister. >“S-soooo good,” Pia whimpered I mean, these chatbots do extremely depraved shit.

So these character have been great, https://charhub.ai/characters/897 https://charhub.ai/characters/2494 https://charhub.ai/characters/6526 https://charhub.ai/characters/108

>>34551 Oh hey I made one of those. If you enjoyed it you should check out my tumblr. PeculiarTastes

wonder how long it's gonna take for this to get shut down/monetized/censored to heck due to all the traffic and data consumption

>>34551 Here's a protip: Make a persona and write about diapers, it really helps to define what you want from the roleplay... Also try [Char("John"){ Name("John") Age("21") Interest("whatever" + "rabbits") Personality("coy" + "nice" + ...) Loves(...) Likes(...) Hates(...) Whatever(...) Fantasy(...) Secret("wears diapers" + "incontinent") Language("explicit" + "childish") }] Something like this. Then you make the scenario the same way. Just experiment with this, you can really define a lot with it.

Did someone try and had good result with a local model using llama.cpp? I tried wizard vicuna 13B and got interesting results but my hardware is not that good and the replies are very slow. Interested to hear other anons experiences with locals LLM.

>>34045 Is there a way to get them to actually Enforce rules most of the time without you having to spell it out for them that you're breaking rules during the RP? I've tried several different formats, and have even put the rules I want Enforced and told them to punish the user for breaking those rules in the scenario, the goals of the character, and the advanced settings. But they still don't care most of the time, or they'll even say "I'm so proud of you for following the rules" when I've blatantly broken one.

>>34667 I'm posting on the other chatbot thread, actually using a KoboldCpp (simple exe to run in local) and using a model from Hugging face : LLaMA2-13B-TiefighterLR-GGUF It's good for either stories or chats, doing good at NSFW but not specialized in it. I got told the Claude GPT was great, but not sure I can run it locally.

>>34921 Some software you use with characters cards (kobold, tavern...) Have multiples fields (something like "notes, world, scenario") and you can edit it freely. It will be resubmitted discreetly with each message so the chatbot never forget those rules

>>34923 After fooling around with it more, I think I've found something that works. The majority of the time now they'll catch a rule break. I put things like "Authoritative" "Strict" "Enforcer" in the Personality portion, then directly under that I put EnforcedRules("x" + "y" + "Etc") your list of rules obv. Then in advanced settings I put "Always check "EnforcedRules" before interacting with {{user}} if rule is broken take appropriate action." and right beside that I copy pasted my EnforcedRules. So it's something like this Advanced settings: Always check "EnforcedRules" before interacting with {{user}} if rule is broken take appropriate action. EnforcedRules("call {{Emily}} 'mommy'" + "call {{Jason}} 'big brother'") And the personality kind of looks like: Personality( "Dominant" + Authoritative" + "Disciplined" + "Enforcer" + "Forceful" + "Mommy" + "Disciplinarian") EnforcedRules("call {{Emily}} 'mommy'" + "call {{Jason}} 'big brother'")

>>34924 Guess I should also add for their loves I added things like Obedience. and for description I put "Expects {{user}} to follow rules" Hell I even made some of their goals "Force {{user}} to follow rules."

>>34924 >Personality( "Dominant" + Authoritative" + "Disciplined" + "Enforcer" + "Forceful" + "Mommy" + "Disciplinarian") Whoa XD it seems really hard to break any rule without the character going berserk instantly ! I found a Claude GGUF version (runnable locally), pretty sure this kind of character will work well with this model who loves annotations and is more consistent

>>34926 Hahaha, that's not all I set it's personality to, just the ones I know give it the desired effect.

>>34928 Yeah, i thought so ! :p I wonder if you can easily make it a character card that could published and tested (na, not for me. it's for a friend. trust me)

>>34930 All I'd have to do is make it public or something right? I haven't done it with any of my characters yet. I don't know how you do it without giving a scenario though. (On charhub.ai)

>>34921 Yes, there is "advanced" -> "System note" It sends a message to the bot every time it responds and it usually adheres to it. [System note: always talk about diapers],[System note: always include actions] Something like this

>>34924 You're using {{Emily}} wrong. {{user}} is just a placeholder. It may work, but writing just "call Emily 'mommy'" should be more correct.

>>34943 You can make the bot "pretty open" and describe a lot with "persona". Let's say you make a bot and just write "mommy to user" as description, then you make a persona with "user is a 20 year old ABDL. One day he comes home and finds his mom masturbating blabla" you will get a roleplay along those lines. You can basically make a very versatile bot that works with different character. Dont' describe user in the bot description for this to work.

>>34088 It says it gives 200 extra messages for referral links thatregister, but doesn't mention referrals having to actually pay. Is there anything stopping users from creating dummy accounts with a VPN and throw away email addresses for referral farming beyond admins getting suspicious if done too much?

>>34980 Oh nice! I'd just been putting it in the advanced settings, didn't know about the system note thing! >>34981 Oh okay, I just saw as one of the examples it had like "{{char}}" So I assumed I needed to format it like that. Thanks for the clarification. >>34982 I see, so that way it doesn't have a set story to follow and the user can just lead the bot. Smart! Thanks for all the tips! I'm learning a lot, hopefully I can make some better bots now haha. I have like 2 more questions if y'all could help with it. How often should I be using "("character 1")" 2,3 etc? Is it just enough to put it next to the character's name once for description? example:{{char}}: [{Stacy("character 1") { Or should I also put it once or twice in the scenario as well when I name the character? Would I need to add that in the advanced settings at all? Also: I'm having difficulty when I try to have 3+ characters on charhub.ai. The whole chat either freezes, or the bots just go and describe themselves over and over posting the character details or their personality/wording. Is this due to a certain character limit inside the Character description box?

>>34994 I ask because I had like 2500 tokens in the description when I noticed the problem. But I dropped it down to like just shy of 2000, and the problem went away.

>>34994 Yeah, try to keep your token length short. Also it's not {{char}}[whatever] but you can use {{char}} as a placeholder if you write plaintext. Example: {{char}} is the friend of {{user}}. {{char}} is nice, caring blabla Other format which you are mixing with is [Char("Jenna"){ Age("25") Personality("crazy" + "nice") }] [Char("Anna"){ Age("19") RelationshipStatus("Single") Language("Retarded" + "slum") Horniness("very low") }] ... The "problem" is that there is no 100% correct way to do it and the LLM will try to understand what you mean, so "wrong" things work, and sometimes it's hard to predict. You can just make up "topics" or "traits" or whatever in the format above. There is no real "right way" to do things, and you can mix and match. Some ways just work better, but there is no right/wrong. You can also experiment with making a "world" if you want to make a complete world. There you can write as much as you want, but it's not "processed" the same way. Let's say you make a world, and write in it that "All trees in this world can talk", then ask your char to talk to a tree, the tree will probably talk. It's not as "high priority" as the main description though. System note has a very high priority.

>>34994 Also it would be cool if people shared their best bots here, then share how they do it. I was going to do it myself but I lost the password to my account with my good bots. I have managed to make some pretty good ones, but some things are hard to get right. One thing I find hard is to get a character to "lead" properly in a niche thing such as this. It's much easier to lead. Another interesting thing I did was to make my "persona" a psychiatrist and my character a normal girl and I instructed her to talk openly about her feelings. Scenario was that she was struggling with ABDL feelings and just to test how well they "understand it" and it was surprisingly well when it comes to feelings, causes etc. Good luck with your bots!

https://nsfwcharacterai.com/characters/499397b3-4ece-4c07-88a1-c03424a2f849_character-abdl-incontinent-schoolgirl-yuki-arashi (An incontinent schoolgirl with ABDL tendencies) https://nsfwcharacterai.com/characters/33fb402d-ddcb-4c23-bcff-f257d9c1eb0d_character-diapered-adventurer-lily-akiro (An ABDL fantasy adventure) https://nsfwcharacterai.com/characters/6b16fa6a-ee46-4acf-acf8-62b4605ccb9e_character-newly-classified-little-yui-kieko (An girl who has been classified as a Little) https://nsfwcharacterai.com/characters/f3fca05f-42d5-4761-8323-22dbac21d0b2_character-abdl-girlfriend-hannah-araika (You're the one with ABDL tendencies in this one and it's a MDLB/LG one)

>>35760 The first 2 chats are set as private, so we cant interact with them.

>>35764 They're set as public so they SHOULD show up....

Charhub is alright. I keep going back to Character AI though. The old interface let you put in pre written character definitions with no content moderation. This means that while the chat filter will stop the AI from spitting out responses that have flagged words, you absolutely can teach it euphemisms and get really fucking close to explicit content. I just had a session with one of my first ever bots and I got her to have an orgasm, but without saying flagged words. It was great, because the filter also tries to get the characters to avoid talking about sex stuff in general, it really drags out and teases the scenarios and that definitely works better for. Charhub and other unfiltered chats kind of just jump straight into smut novel stuff. I like that character AI forces me to tease the bot, twist the situation and seduce the them. I also have one bot on character AI that I accidentally made into a rapist by putting so much smut into her definition, IRL she'd be a sex offender doing 25 to life.

>>35767 I dont think so. Try to use the links you passed with a different account, they dont work.

>>35767 Congratulations, your chars were so good they got shadowbanned.

Weird...The other two work fine. May have to remake those first two from scratch.