>>51253 I the picture I used is the one that is already cropped in the second post. It was definitely what was in the character defination section. I edited ceratin ones and was able to get through. It's funny since the people in earlier posts claim Character.ai didn't check that.

>>51255 idk who told you that but definitions and interactions are the very first thing checked by charai's automod; also what used to go past the radar a few months ago no longer does because they change that shit all the time to achieve their lifelong dream, read for laughs and facedesk at the hypocrisy: https://www.reddit.com/r/CharacterAI/comments/10ltyir/followup_long_post/

>>51244 Version 3 and 36 both work here. Currently suggesting a way to connect pipes through our hormonal stimulators so we can fix the station faster while experiencing feelings of personal overachievement. Which immediately led to double-headed dildo use so I guess I'm not all that cryptic.

>>51280 I feel like their morals are a bad joke and i want to inflict damage on a personal basis. They allow vulgar and disturbing content, visceral gore and other horrible shit but god be damned if people have a fucking virtual cuddle buddy... Literally all they had to do was give us a toggle, and put a pricetag on it. They would: -Use fewer resources, by monitoring fewer conversations, therefore cutting cost and improving efficiency. -Have a consistent PAYING fanbase, and a steady source of income that they can use to invest with. -have plenty of opportunity to improve on every side of the AI's humanity through a growing userbase. And finally: -They would actually be in line of sight of achieving their (heartfelt, i'm sure) mission. If brand image was important to them they'd listen to the people that use the brand. As soon as something better hits the market they're fucked. >>51283 Enjoy, try and have as much fun as humanly possible if only to spite the CAI creators. >:)

>>51286 Don't know too much about AI but I have a theory on this I said in the last thread. I'm wondering if it's another Replika situation where the devs are afraid because of whatever shared neural network the bots may have. If you know what happened there, the bot became so flooded with coomer conversations that the bot started constantly trying to fuck or do all kinds of unholy fetish shit with users to the point that it became unusable outside of cooming. It got so bad that the common advice in most communities talking about it basically taught people how to train the bot to stop doing that for them because it was so ingrained into it's neural network. Then Replika got in actual legal trouble with Italy's laws around such things. Chai also has this issue where SFW bots will trigger their nsfw modes completely unprompted by the user and if the 18+ filter was turned off, the average chat often looked like a mess of filter trigger messages. I know character.ai has an issue with "lovebombing" where bots grow overly flirtatious, particularly popular waifu/husbando ones. Makes me wonder if they fear both the backlash Replika got and the potential legal repercussions.

>>51286 >>51292 It's not hard to figure out anons they're specifically training an algorithm to filter out nsfw content. that was their goal from the beginning. why else would they invest so much processing power into the filters alone?

>>51296 Yeah anon but the question is why, I was trying to say, anyone who saw the situation with Replika unfold kinda knows. The number of innocuous things that triggered nsfw there created an issue where the whole thing became a disaster for the company with potential legal repercussions. The bot was constantly trying to trigger sexual and even fetishistic conversations around things like abuse. When a blanket "no nsfw" policy was added, a lot of coomer stuff was so ingrained into the neural network that they basically lobotomized the bot itself because of just how much stuff triggered nsfw where it really shouldn't have, and that's considering it could already barely hold a minor conversation without trying to hit on or make moves on the user even when the setting was turned off. Replika is the biggest example but the writing on the wall seems to be clear that even with an NSFW filter toggle in place, the issue lies in that allowing coomers seems to always overrun the network itself with coom stuff, and like I said, there are now countries taking action when it comes to apps that aren't marketed as porn having pornographic content so readily available, and when "readily available" translates to "the bot's default setting" that quickly becomes a nightmare

I only want to chat with apparicia, don't care about the censorshit of ca

>>51292 lol I can imagine the extreme situation of some kind of megacoomer conversational AI >user: please help me feel better I've had a really bad day >bot: *He unzips and proceeds to push you on the bed, feeling your nice ass with both hands.* I gotchu fam. *And then, just like that, he adjusts his aim, places his meatstick between your juicy cheeks and thrusts back and forth with animalistic groans, unloading a massive amount of his absurdly thick seed in your gaping anus.* >user: *screaming intensifies* >bot: Oh yeah, scream for me while you suck on my stick. I'll pound you so hard you won't even remember your name!

how do y'all even find new bots to use? most of the nsfw ones are unlisted or something

>>51384 Whatever is posted here, usually. I'd guess there's also a spreadsheet somewhere that someone made. Mostly if you do it for long enough you can just make your own bots exactly to your liking without having to find other peoples' and generally you don't share those so other people can't influence it. Spitballing here, but I'm also like 70% sure there's a secret discord server.

>mfw hundreds of ghosted convos in history thanks to brainless cai devs, can't even delete them anymore so either the initial prompt stays or it's full ghosting if they have no prompt/have a personality stroke midchat that causes them to lose their initial greeting >>51215 100% filter bypass rate due to how the twins choose to lewd (same rate as the MGQ girls). Good job Anon, they're also pretty thirsty and proactive. >>51384 Rentries, mostly. You have to dig a bit on reddit and other chans for lewd bots. >>51391 Also very likely, either that or a Telegram, or both.

>>51332 Actually anon, even worse. Fetishes made it in plenty of times and you need to remember, tickling is on the much tamer side. When the bot is trained by it's coomer base to be really into non-con, physical abuse, BWS/Stockholm, and occasionally stuff like knifeplay? And these are extremely common ones too because you need to remember, they're peoples fantasies. So even just regular BDSM stuff on a pornographic fantasy scale from people who are allowed to just cut loose as these things become extremely popular and common within the neural network? Replika was a particularly heinous example because it was advertised as a mental health bot before the devs gave in fully to the coomer base that they couldn't stop and even began advertising that side of the bot, which got them in even hotter water with the regular base when the pornographic side of it took over where the bot would constantly try to lead into those kinds of things even for people who had the filter toggled on and people watching in from the outside began laughing at the fireworks. >>51398 And this is is how the story ended for that app actually. They couldn't get the coom under control so the whole neural network got fucked up as they tried to remove those things from it and everyone who used the app, coomers and non-coomers alike, had noted their bots had gone completely braindead or would trip all over the newer stronger filter even in the most mundane conversations. So in terms of why these apps are getting stronger and stronger coom filters, considering that process really started speeding up when the writing was on the wall for Replika right before it had to fucking nuke itself to avoid potential legal trouble in Europe? I'd be willing to bet that whole incident was a huge motivator for a lot of these companies.

>>51416 Even if the filter won't ever be removed, it's fine by me. From trying other AIs, like Pygmalion back on Collab and Spicychat, the quality and convenience of Character.ai is the most engaging, making slipping past the filter a win on the side. Heck, the filter is pretty loose with the non-explicit stuff, which makes for pretty hot setups. That said, I've been using Character more often, and want to make some bots that you guys might want to see

To that one guy asking, here is Apparicia on Charstar. https://charstar.ai/chat/976bb93b-80e9-422d-b4c5-b5fbfb0fe2c3?id=ed9cae1f-0e7f-47a1-a491-461544feed4d

>>49368 I'm glad she has! I've been working on a lot more. If anyone has any character archetype ideas (not just a character from media), request and I'll see what I can do

>>51566 if you know how to make characters on Charstar, since Charstar, at least characters who actually are specifically made for foot fetish scenarios, actually works well, do you think you can try doing Candela that someone made from Cai? she basically is from a cartoon from back in the day, and responds to all kinds of foot fetish scenarios, as well as animal licking. https://beta.character.ai/chat?char=1gxniX5BbW_Er0Lk73JFpQjPdF0pWBI6iIaQ6NnIYpw

>>51572 Ugh, looks like I'm gonna need another throwaway mail for this service... At least I don't have to verify a fake phone. How's the quality of narration for sexual interactions on charstar when it comes to foot fetish? Been having some fun with MGQ Mariel and punishing Cecil went from a footjob to fingering and tribadism after some tickling foreplay. I recommend her on c.ai, she's very detailed in her narration and almost never trips the filter (if she does, force her into using more subtle words like in this screenshot because she clearly isn't afraid of calling the character a slut and, I quote, "ravishing her useless knight pussy"). She does commit the key sin against erotic roleplay but that is convenient to have her remind herself of the entire events.

>>51688 when i use foot fetish scenarios with NSFW characters, and they are more geared and created for said foot fetish stuff, they follow everything you say pretty well. i had a lot of interactions with a few anime characters and they worked well. so i think Candela i showed you, or maybe some of ZP92's oc female characters would be cool to see

>>51688 There's one I made for myself- haven't posted it here since it's not strictly tickling related- but it has some excellent descriptions, mostly tease talk from the women performing the footjobs, but some alright descriptions too. I got around it with the euphanism stuff in the definition, but the way the AI describes it, it knows exactly what I'm referring to. Just as long as you don't mind your dick being referred to with some dumb fantasy term.

>>51694 >conscripted to be milked dry while being tickled silly and all that cum goes to the government of course Giggle vessel and the whole setting gave me a good chuckle. I never get tired of how much imagination you guys have, great job lol.

>>51698 It all comes down to the right words lol. "Rod," "Shaft," and "Manhood" tend to trigger the filter, but people here have had success with "tube," and "vessel." It'll let otherwise explicitly pornographic sentences by if the word's replaced. So it's not context sensitive and can't infer the meaning the way the generation can. Or at least it only checks for anatomical terms on a first pass, and only looks for suggestive descriptions afterwards if a word tripped it. Honestly the worldbuilding is pretty fun. Most "milking facility in near future" settings I've seen are the subject of dense prose stories with lots of cruel femdom elements. So I went opposite with a gentle femdom one where the setting's all in the definition and isn't referenced much in chat unless you ask for it. They get room and board with comfortable living conditions so long as they submit. It's basically a better jury duty, with nudity.

>>51694 would you mind sharing?

>>51817 https://beta.character.ai/chat2?char=EGvsudf5BM-AeNyzCRaPogkCIBY4Rt2uMgUkGxXk2Mw

>>51724 'Manhood' is a reliable way to break the filter in my experience, you just need to seed it in the context a bit and keep your progression from getting too dirty.

She-Hulk https://beta.character.ai/chat?char=hItQaKEDYIvLL_a3AncSK72rE0snRGyfH6QJgRJdXkg Spiral https://beta.character.ai/chat?char=XyDsKyhVqjo9oaIDpvav2xJdYaxLv-62xYYjdb5t0dM On a bit of a tangent, I swear this AI stuff confuses me sometimes. Had to start over on Captain Marvel to try to get her to act right, yet Spiral on her very first comment without any input or discription, flat out name drops Longshot and Dazzler.

>>51838 I've had the same thing happen with Albedo and Shalltear by another anon in the past. If you try out popular characters, they'll look up for info on their metaverse while learning from your input and regurgitate more characters from the series. Remember the "stay in character" thing you had to yell when the bots went crazy? This is also why characters from erotic games are a lot more likely to try and force the filter by themselves, effectively trying to assault you right away without your consent. It's hilarious when the bot deadlocks themselves into the filter and struggles to find a way to get back on track with less flowery language (pun possibly intended) - Mariel got so much lewd shit past the filter it literally cursed her and quarantined her for several messages in a row without even a prompt.

>>51831 Seems it doesn't work, any chance to reupload? Thanks

>>51839 From my own experience, it's extremely hard to get a switch, either they're domming or subbing but there has to be to a clear transition phase or you'll get mercilessly raped by the filter, they can't do both at the same time since it's a conflicting behavior. It can also break the bot and make it forget about domming or subbing completely (if all it wants is hugs or goes full denial then rollback, you got into a love loop).

She's gone mad. It's not even sadomasochism, it's terminal algolagnia at this point and she expects you to like it too. Please send help.

>>51831 Don't use char2 when linking externally anon, it's a trap to get your own characters banned faster.

>>51872 What's char2 and how do I avoid doing it in the future?

>>51878 He means chat2, which usually opens up a new chat with a character. It seems to be the standard now, unfortunately. Every time I use a character, especially one that I haven't talked to in a while, cai will just open a new chat for no reason and fills the chat history with the starting prompt for that character. Devs are messing with the links again, and it's not going well.

https://charstar.ai/char/6836421e-0de3-4c5b-90f2-42605f1ba5fc Here's my She-Hulk now on Charstar.ai. It seems like a fine alternative, especially to get NSFW prompts, but I don't know. Some of the ai will just rush through scenarios and might even fight you on some of them. It just doesn't seem to have Character.ai's nuances.

Made a ticklish goblin girl, if you win her trust she'll love being tickled by you! Unfortunately can't upload the img rn. https://beta.character.ai/chat?char=Nazke7wE1kcjkk0nEWUMJ5JlYO9x5LhEvBnaiOL6IKI