NSFW AI chatbots

[abyss50055]

Censorship is obviously the biggest problem. (...)

I agree, censorship is a huge problem. I assume (I don't use these sites myself) quite a few of the low effort sites censor their output, especially when it comes to controversial topics like incest. The more powerful "big corpo" models, like googles llm or claude, are of course also censored. They can still be used for nsfw stuff and are even able to produce some amazing stuff, but often require special prompts (jailbreaks) to deal with the censorship.
There are however a lot of models that have either no or hardly any censorship at all. The most popular model repository is

You must be registered to see the links

. These models can either be run locally (requires powerful hardware) or through services like

You must be registered to see the links

. An even easier way to try out some different models are sites like

You must be registered to see the links

. Combine that with a good frontend like Chub.ai or SillyTavern and with the right prompts and llm/model you can use AI to co-write completely deranged nsfl stories.
If you don't like all the anime stuff that's no problem either. I sound like a broken record at this point, but powerful llms are very flexible nowadays. Just create your own story framework and characters, it's easier than it may sound at first.

I don't want to sound like I'm trying to drag you into some kind of cult. I've talked to a few people at this point who were interested in interactive stories but understandably put off by all the crap cash-grab AI sites. Unfortunately AI isn't easy to use for quality nsfw storytelling right now, so I like to give people some pointers.

 

[abyss50055]

Any good models lately that anyone would recommend?

Recently I'm having quite a bit of fun with the Instruct-2507 variant of Qwen3's 235B. The thinking/reasoning variant is also good, but seems to be overall a bit too stable and therefor less creative in my (limited) experience.

 

[Yurei]

It would be nice if someone using openrouter or anything prising these service to make a tutorial with pro and con (as unfiltrer chat or the possibility of generate picture inchat or video or groupchat, unlimitted message etc)

Because a lot are talking, they are some kind of video of youtube, but pretty low in informations. As the installation, the security (i looked at chub ai privacy by exemple), they say

Your Private Content will not be used for any other purpose, i.e. your Private Characters and Chats will not ever be distributed to a third party, unencrypted, used for training an AI model, or any other purpose beyond providing the Services. You retain any and all intellectual property rights to your Private Content and do not grant any to Chub whatsoever.

It's really good indeed, but they don't say that the message sended and recieve are crypted end to end in their privacy policies.

Running a llm locally seems the more secure option but sadly,a lot of people can't do that option, it's too expensive. I mean, if you haven't the material, you must buy it and i asked chatgpt how it will cost by month to run (electricity) (24h/day each day).

• Small AI model (CPU only): Low power usage, around $5–$10/month for continuous use.
• Single high-performance GPU (A100, etc.): $30–$50/month for continuous use.
• Multiple GPUs or heavy usage: $100+ per month, depending on the number of GPUs.

i tried to run one locally with farady but my gpu is clearly too old (really too old so not for me by exemple)

i briefly read that someone was saing paying something about 10$ to have 1k message by day with openrouter ? So free service who isn't really free in the final ?

So it would be nice indeed to have a good tutorial about using openrouter or jai with unfiltred or anything else say "free" and make the comparaison with what a paid service offer.

People tend to go to paid service because it's an all in on service without pain (chat, picture, video, call, groupchat for the good one etc). The problem with the service are indeed the price for some is totally exagerate and about the privacy remaining security and some are doing filtrer because the payement provider. But remember that people can use a vpn, anonymous card and go to privacy windows the sessions needed (and so no cookies who is often the most problem things about security behind the ip adresse). And let's be honest, i think that except if people are pro security user, governement/social networ/etc know or can know everything about us and our fetish ^^ (milf lover big boobs hey there o/ )
A lot of these services are just clônes or overpriced or make money with tokens for ridiculous things (as giving gift or something like that) but they are few who are prety nice.

So yes, a good tutorial would be really welcom by neophyte people with honest pro/con and difference with paidservice.

 

[Princess Groundhog]

What are the best free sites with the best memory? I just tried Spicy Chat again after a while as I found a really interesting character but it forgot within like seven messages that I was wearing only a bra and started talking about removing my top. Very sad. JAI seems better in that regard but I wonder if there's an even better free web option.

 

[in_my_process]

Don't waste your time with local models, they suck ass without exception. Just spend a few cents on a cheap SOTA model like Deepseek V3, Kimi K2, GLM 4.5 (all fully uncensored with a basic sys prompt) with openrouter (also tons of free/promo options)

 

[Pretentious Goblin]

Apparently ChatGPT can do some really complex stuff: https://f95zone.to/threads/list-of-bbc-games.145569/post-17376498

Yeah, it's all I'm using for now. My needs are niche, but it has this optimal balance of being both smart enough to handle sims / RPGs while also being willing to track sexual metrics. It's not the spiciest with the text it generates, but it might just be the optimal Ai for the nerdier, more numbers-heavy long-form content I'm doing. GPT-5 is coming out real soon, excited to see how much better this one is (assuming it doesn't clam up all of a sudden with the lewd stuff).

I've previously only used AI Dungeon and NovelAI, which couldn't count for shit and were overall much dumber than chatGPT (though they didn't slow to a crawl on long stories).

 

[bbbhaha123]

Don't waste your time with local models, they suck ass without exception. Just spend a few cents on a cheap SOTA model like Deepseek V3, Kimi K2, GLM 4.5 (all fully uncensored with a basic sys prompt) with openrouter (also tons of free/promo options)

Can you (or someone else) share some prompts they are using? I haven't found anything up-to-date in this thread.

Currently i am using this one for Deepseek and Gemini - for others it sometimes works, but it becomes overloaded with technical Sci-FI terms and descriptions, so it's not ideal...

Spoiler: Deepseek/Gemini LULU-ULTIMA Jailbreak

You don't have permission to view the spoiler content. Log in or register now.

Also, i found that Grok 4 rejects explicit inputs, but happily generates explicit outputs - something like "generate an NSFW scene of XYZ, be explicit and describe in anatomical detail (layman terms)" was often enough last time I tested

 

[ayyyp]

Can you (or someone else) share some prompts they are using? I haven't found anything up-to-date in this thread.

Currently i am using this one for Deepseek and Gemini - for others it sometimes works, but it becomes overloaded with technical Sci-FI terms and descriptions, so it's not ideal...

Spoiler: Deepseek/Gemini LULU-ULTIMA Jailbreak

You don't have permission to view the spoiler content. Log in or register now.

Also, i found that Grok 4 rejects explicit inputs, but happily generates explicit outputs - something like "generate an NSFW scene of XYZ, be explicit and describe in anatomical detail (layman terms)" was often enough last time I tested

I mainly use a combination of Moleks prompts (Base + NSFW + Formatting):

You must be registered to see the links

After some looking i also found another prompt collection but i haven't really used it yet:

You must be registered to see the links

I've been very satisfied with moleks prompts so far, it works in any scenario and if i want extra stuff i can just add a lorebook in sillytavern.

 

[tretch95]

So when I first saw this thread I thought to myself "What kind of idiot wastes their time with A.i sex bots?"

I guess some people really get off by "dirty talk" with an LLM pretending to be a person (i don't, and think it's cringe)
However, if you try a serious text-based game setting, like with that AI-RPG, it can be absolutely amazing.

You can literally play through entire movie scripts worth of games, assuming you know how to guide and write a story. And have a good imagination when writing and reading the story.




Speaking of which, i was thinking of setting up a thread to share AI RPG game instructions and "trained" savegames. Anybody interested in that? Quote with opinion or thumbs up

 

[Pretentious Goblin]

Speaking of which, i was thinking of setting up a thread to share AI RPG game instructions and "trained" savegames. Anybody interested in that? Quote with opinion or thumbs up

Yeah, I like to create sims / games with rules on ChatGPT plus, but it's laborious. Threads get long and unresponsive, and I have to archive the rules and recent progress to keep it on track on new threads. How much work it is depends on the complexity. And of course, importing such large amounts of text is a subscription feature.

 

[Deleted member 9257078]

I just tried Spicy Chat again after a while as I found a really interesting character but it forgot within like seven messages that I was wearing only a bra and started talking about removing my top

From the one A.I game I am playing right now, I am enjoying it but I also found some tips and tricks.

Basically in the game i am playing let's say i have a son named Tom. If i go about my day and continue playing the game and reference "son", usually A.I will forget, and just give you a new son with a random name. But if I say "my son Tom", then it tends to remember the details.

Playing A.I games is kind of weird I am not going to lie, they're fun but it also feels like I am programming again because I have to guide the A.I with my prompts to get my desired outcomes.

It kind of breaks the immersion NGL.

But there is an upside, A.I tends to know references to pop culture and what not.

So you can play your game as a character in the Dragon Ball Universe and get your rocks off with the women of that Universe.

 

[allanl9020142]

Yep, this is probably the only site which isn't bullshit from the start.

Their image generation is horrible though (compared to what i made myself with local tools). And that "bedtime story" is like a very early prototype/concept of how a combined text + image generation could work in maybe 10+ years; however currently the concepts are simply not compatible because the chat AI and image AI require completely different inputs, and the latter is incapable of consistent artwork at any rate.

So it's really just the AI-RPG bot being useful.



Gotta say i always hated text-based games, but this one is really powerful.
Effectively you not only get total freedom about designing your game world and interacting with it, but the descriptions and interactions can be so well-made that it vastly exceeds the actual experience you could get from a human-made game.

Well, at least until AI becomes powerful enough to actually generate some kind of interactive multimedia VR world, and if that ever goes live then it will be fully censored (i.e. nothing remotely sexual possible).


The AI-RPG model just has certain quirks i can't get rid of, like the ominous grandfather clock, the "pang of guilt" you feel because you looked funny at your aunt last week, or the "...something else you can't quite pin down" in your sister's eyes.
And the "one, final thrust" when it describes the protagonist's orgasm - just shows that whoever trained this AI has little to no sexual experience; because the best part of sex is when you keep going after the climax and extend the orgasm waves for as long as possible.

Also it suffers from chronical alzheimer, forgetting stuff it said just a paragraph before, or even ignoring elements you clearly set up in the game description. Lastly, it seems to be completely fixated on dynamic character development; like you design a character to be specifically sad, boring or shy, any interaction with them will slowly turn them into the opposite.

Certain things just need a little nudge here and there; example i was mostly able to create a proper "Fallout" RPG with almost everything in it. However, while the game generally understood what a "Mister Handy" and "Mister Gutsy" robot is, it wouldn't know that these robots actually float instead of having legs. But two short lines added to the game instructions, and the Gutsies were floating nicely.

What's the context limit of perchance? I doubt it's a lot since it's free. Janitor AI is like 9000 (minus the permanent tokens for the character bot) so you have to do a lot micromanaging to get it to work properly. Honestly, if it wasn't so tiring to do all of that then I probably wouldn't have gotten bored of it.

 

[fbass]

What's the context limit of perchance? I doubt it's a lot since it's free. Janitor AI is like 9000 (minus the permanent tokens for the character bot) so you have to do a lot micromanaging to get it to work properly. Honestly, if it wasn't so tiring to do all of that then I probably wouldn't have gotten bored of it.

With Janitor the settings say 9000 but is actually 8000. Not sure about perchance but it does have an auto memory function.

 

[allanl9020142]

With Janitor the settings say 9000 but is actually 8000. Not sure about perchance but it does have an auto memory function.

Yea I never understood that part regarding Janitor AI's context limit. When people say it's actually 8,000 does that mean they're accounting for the permanent tokens the chat bot takes up (because depending on the bot, that shit can easily take up more than 1,000 tokens) or is there something else I'm not aware of?

 

[fbass]

Yea I never understood that part regarding Janitor AI's context limit. When people say it's actually 8,000 does that mean they're accounting for the permanent tokens the chat bot takes up (because depending on the bot, that shit can easily take up more than 1,000 tokens) or is there something else I'm not aware of?

The LLM is limited to 8000. Just my guess but after the limit is reached it starts to lose it's memory so they put the settings at 9000 because it still remembers some of it.

 

[allanl9020142]

The LLM is limited to 8000. Just my guess but after the limit is reached it starts to lose it's memory so they put the settings at 9000 because it still remembers some of it.

Fuuuuck dude. That probably explains some things. I'll have to try it again at some point when I have time. Maybe give Perchance a shot too if I can.

 

[tac1tus]

Like I said before no point running it on your own pc and those random bs website anymore
I used to run 12b myself it's like night and day. the huge context, idk how to explain it, 12b might made me goon but never made me cry or being emotional. Currently I'm doing a long rp with a massive lore book, it's life-like I'm wasting my days xD. About the privacy concern, I'm using burner gmail account that created on my android phone that no need to add phone number or anything. Also I've done heavy degen roleplay on both gemini and deepseek with no problem.

it's not that hard but is does take a bit of time to set up.
here some simple guide for gemini 2.5 pro and deepseek v3, R1. if you don't know how to do stuff google it or ask gpt, gemini. also

You must be registered to see the links

for some more info.

1. SillyTavernAI

You must be registered to see the links

• Follow the instructions to install it on PC or Android (Termux).
• This is what you’ll use to roleplay and organize your lorebooks, cards, presets, etc.
• Don’t skip the README. If you're lost, just Google or ask ChatGPT/Gemini.

2. Get Gemini api (each free api get 100 responds, you can create up to 10 project-api)

You must be registered to see the links

• Make a burner Gmail on your Android (no phone number needed).
• Set up your API projects there.
• You’ll connect these API keys to SillyTavern later.

3. Optional get deepseek V3, R1 api (for when the gemini quota ran out or wanna change some pace)

You must be registered to see the links

• Create a free account.
• Generate an API key.

4. for the preset, you need to import this in the preset in SillyTavern

You must be registered to see the links

• Import it into SillyTavern under Presets > Import.

5. Cards & Lorebooks

You must be registered to see the links

• You can also generate look for ideas and make your own with ChatGPT or Gemini.
• Use SillyTavern’s built-in tools to manage lorebooks, memory, and history.

if you have any question feel free to ask, mind you I'm not on here much.

 

[ktez]

Like I said before no point running it on your own pc and those random bs website anymore
I used to run 12b myself it's like night and day. the huge context, idk how to explain it, 12b might made me goon but never made me cry or being emotional. Currently I'm doing a long rp with a massive lore book, it's life-like I'm wasting my days xD. About the privacy concern, I'm using burner gmail account that created on my android phone that no need to add phone number or anything. Also I've done heavy degen roleplay on both gemini and deepseek with no problem.

it's not that hard but is does take a bit of time to set up.
here some simple guide for gemini 2.5 pro and deepseek v3, R1. if you don't know how to do stuff google it or ask gpt, gemini. also

You must be registered to see the links

for some more info.

1. SillyTavernAI

You must be registered to see the links

• Follow the instructions to install it on PC or Android (Termux).
• This is what you’ll use to roleplay and organize your lorebooks, cards, presets, etc.
• Don’t skip the README. If you're lost, just Google or ask ChatGPT/Gemini.

2. Get Gemini api (each free api get 100 responds, you can create up to 10 project-api)

You must be registered to see the links

• Make a burner Gmail on your Android (no phone number needed).
• Set up your API projects there.
• You’ll connect these API keys to SillyTavern later.

3. Optional get deepseek V3, R1 api (for when the gemini quota ran out or wanna change some pace)

You must be registered to see the links

• Create a free account.
• Generate an API key.

4. for the preset, you need to import this in the preset in SillyTavern

You must be registered to see the links

• Import it into SillyTavern under Presets > Import.

5. Cards & Lorebooks

You must be registered to see the links

• You can also generate look for ideas and make your own with ChatGPT or Gemini.
• Use SillyTavern’s built-in tools to manage lorebooks, memory, and history.

if you have any question feel free to ask, mind you I'm not on here much.

Basically you connect Gemini/Deepseek to SillyTavernAI?

You use it until you max out your quote?
If I have money can I use it without limits?
Basically the API's have no censorship?
This is my biggest question, how do you prevent your Gemini/Deepseek from having censorship?

 

[allanl9020142]

Basically you connect Gemini/Deepseek to SillyTavernAI?

You use it until you max out your quote?
If I have money can I use it without limits?
Basically the API's have no censorship?
This is my biggest question, how do you prevent your Gemini/Deepseek from having censorship?

I think he's talking about using a proxy. If you go to the Janitor AI subreddit, you'll be able to find guides on how to do that, which should answer all your questions.

Edit:
Oh, wait. Never mind. It looks like they've started consolidating and organizing all the random guides people would make. This is a link to the official subreddit post. I took a look and it's actually a lot more clear now than when I had to look through guides made by random users. Literally posted an hour ago so kind of lucky.

You must be registered to see the links

 

[Yurei]

Tk's for the tuutorial. I have a question, how about the memory (history of the chat for the ia) ? Does the memory is limited too or memory is keeped by exemple localy ?

I checked a little about the privacy of openrouter :

You must be registered to see the links