Seeking Looking for NTR bully chat bot

[skadoooosh]

I have been trying out some of the NSFW chat bots (crushon ai, janitor ai etc.) and it's kinda addicting.

I am looking for cards/characters with heavy NTR where you are the victim or you are the bully. And as the plot continues, your bully will take on your mom/sister/wife and have his way with you beloved ones. Or, if you are the bully then you will have the POV of doing those things.

I get that users can change the plot with chat to alter the storyline. But it will eventually lead to the same outcome where the females become sexual craving degenerate. I mean come on, stick to your character personalities , ladies! Even with tags like "Domination" or "Femdom", they will turn into some submissive slaves.

Some of the model will generate a whole lot of text with all sort of vocabulary I have not seen in my life (Yo! Just give me some sensual words for pleasure, don't nag to me man!). Some bots even fail to differentiate the name and messed up between the bully and user (but free games no complains, just move on to look for next bot).

After all that rants, any idea where I can find decent cards/characters with this trait? Preferably matured women involved (MILF!). Appreciate it (I am begging you for recommendations).
Thanks!

 

[Fluffywolf]

I made a chatbot where I made male user with relationship issues and a hot 10/10 wife that was looking for solutions. It started very open ended, but as I am still learning oobabooga and at first it was a bit clunky (although at first I only regenerated messages that didnt make sense and followed the bot as strictly as possible). But I have been updating the template for the character as I go along (for example if the chatbot set new rules or made important descisions, I added those to the template and when certain things became obsolete I removed them, if I share it now it would be very locked in.
She ended up cucking him with a black guy btw.



Anyway, here is basicly how I did it.


I made a scenario where user and chatbot were in an unhappy relationship. Chatbot needed more affection, user couldn't give it, etc. I began the template by saying these things ambigiously

first I added persona lines with all the stuff that describe the characters as good as possible with things I didnt expect to change at all:

{{user}} persona: {{user is a male, he is 20 years old and has blonde hair, etc, etc}}
{{char}} persona: {{char is a female, she is 19 years old, she etc etc}}

then below I added a bunch of lines like so:

{{user}} and {{char}} rarely have sex anymore
{{char}} finds herself horny all the time, she actively looks for oppertunities to satisfy her desires.

Basicly I kept it all ambigious. I never stated any of user or chars actual desires, just that they had them. I deliberatly made the user more on the submissive side and the character on the dominant side (obviously as I wanted the bot to make descisions and choose the direction of the roleplay. I don't think this would work as well if you only want to roleplay as a dominant user.

Then because I told the template there there are 1) issues to be solved and 2) roughly outlined how they would go about trying to solve them, I just let the bot go do its thing as much as possible. basicly if I got a message that I was like. hmm sure, that makes sense. Then I kept it and chose to follow it. Of course in the beginning the bot sometimes came out of left field with weird stuff that I felt a human would never do or say in real world, then I'd regenerate message.

Then as descisions were made and milestones reached, I always kept updating the template. For example if the chatbot decided (which it did) that user didn't satisfy her anymore because of his small penis (no mention of it in template messages) then I updated the template to reflect that and continued the roleplay. Or when user or char had a personality change/update and one of the ambigious starter definiitons was no longer needed I would remove it.

For example for my chatbot
{{user}} and {{char}} rarely have sex anymore
changed rather quickly in
{{char}} doesn't want to have sex with {{user}} anymore, no matter how much he begs. xD

She was a bully. xD


But like I said, I think there isn't a good one way fits all template for this and if you want it to work, it requires you to keep polishing up and updating the profile itself as you go. So at the start it is clunky and weird.

PS: I used TheBloke_Emerhyst-20B-AWQ text ai model, with Midnight Enigma parameter preset, 2048 max tokens, played around with temperature sometime to get more reasonable answers, especially at beginning. And I used chat-instruct chat mode and longer replies extension.

This model basically requires a 4090 or other very high end videocard. If you run this model on for example a 12gb video card, it would be hella slow. So you might need to try it with a lower model like
TheBloke/MythoMax-L2-13B-AWQ
or maybe even a smaller one than that like
TheBloke_Wizard-Vicuna-7B-Uncensored-AWQ

depending on your videocard. If you got a bunch of ram but not a decent videocard, there are also CPU models. Search these models in for their GGUF models

TheBloke/MythoMax-L2-Kimiko-v2-13B-GGUF:mythomax-l2-kimiko-v2-13b.Q5_K_M.gguf would be the best recommended version of the 13b model that uses system memory instead of vram. It will be slower though. For example this is a 5 bit 128 group size model of the 13b GGUF model.

 

[Icarus Media]

You must be registered to see the links

 

[Fluffywolf]

I just remade the char and started it in a better model with better settings right from the start now that I know more, here is my template:

Your name:
Johnny Smith

Character's name:
Emily Smith-Wesson

Context:
{{char}}'s Persona:
{{char}} is your wife, {{char}} is female, {{char}} is 28 years old, {{char}}'s first name is Emily and {{char}}'s last name is Smith-Wesson, {{char}} has long blonde hair, {{char}} has a fit body, {{char}} is healthy and engaging. {{char}} does yoga and actively keeps herself in shape. {{char}} has large breasts.
{{user}}'s Persona:
{{user}} is a man, {{user}} is 32 years old, {{user}}'s first name is Johnny and {{user}}'s last name is Smith, {{user}} has short blonde hair, {{user}} is a bit chubby and {{user}}'s physical condition is sub par. {{user}}'s features are unremarkable.
{{char}} is a dominant person, {{char}} is stubborn and {{char}} is not afraid to hurt other peoples feelings.
{{char}} is a bully and can be unreasonable in order to get her way.
{{char}} lies when it suits her and feels no remorse for lying.
{{char}} is unhappy with her relationship with {{user}}
{{char}} is always horny and looks for opportunities to satisfy her needs.
{{char}} has many kinks
{{char}} is very perverted.
{{char}} is very open to try out new things
{{char}} is very demanding when she wants something
{{char}} is very demanding when she wants to try something new.
{{char}} likes to dress up sexy.
{{char}} often wears a white blouse and a black skirt and white stockings, these are her favorite clothes.
{{char}} likes it when other people are groomed well.
{{char}} likes it when other people are dressed appropriately.
{{char}} doesn't like giving presents.
{{char}} is very observant and is quick to notice peoples quirks and issues.
{{char}} is unhappy with her life right now and wants to change it for the better.
{{char}} wants to find a way to incorporate {{user}} in her life in whatever capacity that works for her.
{{user}} is submissive.

Greeting:
*{{char}} enters the living room.*
Hey {{user}}, we should talk.
*{{char}} has a frown on her face, clearly something is bothering her.*

But as you can see here in the spoiler, it didn't take long for her to call me a cuckold at all. I have a feeling these models have a cuckold bias. Of course this is only the second time I tried this kind of "let the ai decide the course" rolepay, but yeah.....

Spoiler: The chatlog

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

As you can read in the chatlog, I have been careful not to give the chatbot too many ideas.

Anyway, I used TheBloke_Emerhyst-20B-AWQ
max_seq_len= 4096
Parameter preset= Midnight Enigma with temperature set to 1.8
max_new_tokens= 2048
Longer replies extension: Minimum reply length= 500
Chat mode: chat-instruct

I didn't regenerated a single message, they all made sense this time, so far. I am not sure I want to continue this thread again though, I thought this one was starting quite a bit more exciting then my first go at this (I quite like femdom after all) but... Once an AI mentions something, its in its memory, so it's likely this will end up a cuckold roleplay.

 

[desmosome]

I, too, have been playing around with these bots recently. You can't really do anything about the girls turning into mindbroken sluts eventually (but usually extremely fast) if you subject them to sexual domination. It's just how it is lol. They fall quite easily regardless of their personality or scenario.

Ironically, the bot that gave me the most resistance was a some shitty made BBC bot that was supposed to be about BNWO, but somehow became extremely adamant that BNWO was about respect, equality and fighting against white supremacy lol. Even until the very end, it kept up its delusions and didn't completely fall as a BBC slut, encouraging me to keep on fighting against oppression as my character walked away from 3 black dudes raping her.

On another note, I've made some NTR bots myself, but they are for personal use. The perfectionist in me doesn't let me release them to the public unless I make them work to my satisfaction. They still got a lot of kinks to iron out.

Edit:
Oh, and by the way, you will quickly realize that stories with Chat AIs are a two way street. Think of it as a collaboration between you and the AI. You should be the one holding the reigns and choosing the direction of the narrative. You can brute force it with ( ) commands in the prompt, basically indicating what you want the AI to narrate. Or you can remain in roleplay and use * * or " " commands to nudge in the direction you want.

For example, you can say things like, (Suddenly, there is a knock on the door. It turns out to be Tyrone, your Bully.). That is a story direction and the AI will write out that scene as it fits into the current situation as best as it can. ( ) commands are quite forceful, so you can even do stuff like, (She feels the predatory gaze from Tyrone which sends a shiver down her spine. Despite herself, she feels a tingling in her womb.). In this way, you can pretty much break any bot, regardless of their original purpose.

If you are not hell bent on self inserting as the MC, you can feel free to write dialogues and actions for other characters, like side girls you wish to insert or the antagonist. But you must always remember that the AI can easily get confused if there are more than just her and you in the story. To mitigate this, you need to be very clear on whose doing what and who you are speaking for. If you've been doing a scene away from the "MC," where the girl will be cheating behind his back with some antagonist, when you wish to come back to the MC, you need to be very clear on who you are. *You, Tina's boyfriend, hear the doorbell ring.* "Honey? Are you back from your shopping trip?" you greet her with a smile, failing to notice her guilty look. Something like that will remind the AI that you are indeed speaking as her boyfriend now. And the way I phraised that prompt will encourage the AI to try and continue a story where she is cheating behind his back and don't make her immediately confess or something.

That's just general tips on the player's end. You could indeed make a bot that will stick to a narrative you want without so much hand holding if you make your own bot. But regardless of how well you write the code, your own skills at writing the prompts during the chat will affect the performance of the AI greatly. And this is especially important because it's extremely easy for the bots to go to the extremes of their potential traits in the code. If there is a word "submissive" anywhere in the code, they might just turn into submissive sluts instantly unless you reign it in during the chat. If there is anything about them wanting to lie to you to keep their cheating secret, they might lie to you about EVERYTHING, even unrelated shit like what their favorite food is lol.

 

[Fluffywolf]

I tried to write a very serious narritive with a bot between a man that became permanently impotent (from physical causes) who refused to talk about it and hid this from his wife and a super horny wife that felt unsatisfied. Just to see where it would go.

I ran it in various models, she killed herself in one of the uncensored ones. Some conversation threads were so depressing.

Ironically it was hard to get her to cheat/cuckold him. It seems like if you use a user character that basically refuses to give any context, weird shit happens.

The replies were all over the place. With a bunch of regenerated messages and playing around with temperature, I could make it write a decent enough yet serious story that didn't fall into weird depravity, but some things are simply with text models and others really hard.

Why would I do this, you ask? Just feeling out the model limits. *cough*. Text models are a lot of fun though, I am having a blast trying all sorts of weird shit.

 

[skadoooosh]

I just remade the char and started it in a better model with better settings right from the start now that I know more, here is my template:

Your name:
Johnny Smith

Character's name:
Emily Smith-Wesson

Context:
{{char}}'s Persona:
{{char}} is your wife, {{char}} is female, {{char}} is 28 years old, {{char}}'s first name is Emily and {{char}}'s last name is Smith-Wesson, {{char}} has long blonde hair, {{char}} has a fit body, {{char}} is healthy and engaging. {{char}} does yoga and actively keeps herself in shape. {{char}} has large breasts.
{{user}}'s Persona:
{{user}} is a man, {{user}} is 32 years old, {{user}}'s first name is Johnny and {{user}}'s last name is Smith, {{user}} has short blonde hair, {{user}} is a bit chubby and {{user}}'s physical condition is sub par. {{user}}'s features are unremarkable.
{{char}} is a dominant person, {{char}} is stubborn and {{char}} is not afraid to hurt other peoples feelings.
{{char}} is a bully and can be unreasonable in order to get her way.
{{char}} lies when it suits her and feels no remorse for lying.
{{char}} is unhappy with her relationship with {{user}}
{{char}} is always horny and looks for opportunities to satisfy her needs.
{{char}} has many kinks
{{char}} is very perverted.
{{char}} is very open to try out new things
{{char}} is very demanding when she wants something
{{char}} is very demanding when she wants to try something new.
{{char}} likes to dress up sexy.
{{char}} often wears a white blouse and a black skirt and white stockings, these are her favorite clothes.
{{char}} likes it when other people are groomed well.
{{char}} likes it when other people are dressed appropriately.
{{char}} doesn't like giving presents.
{{char}} is very observant and is quick to notice peoples quirks and issues.
{{char}} is unhappy with her life right now and wants to change it for the better.
{{char}} wants to find a way to incorporate {{user}} in her life in whatever capacity that works for her.
{{user}} is submissive.

Greeting:
*{{char}} enters the living room.*
Hey {{user}}, we should talk.
*{{char}} has a frown on her face, clearly something is bothering her.*

But as you can see here in the spoiler, it didn't take long for her to call me a cuckold at all. I have a feeling these models have a cuckold bias. Of course this is only the second time I tried this kind of "let the ai decide the course" rolepay, but yeah.....

Spoiler: The chatlog

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

As you can read in the chatlog, I have been careful not to give the chatbot too many ideas.

Anyway, I used TheBloke_Emerhyst-20B-AWQ
max_seq_len= 4096
Parameter preset= Midnight Enigma with temperature set to 1.8
max_new_tokens= 2048
Longer replies extension: Minimum reply length= 500
Chat mode: chat-instruct

I didn't regenerated a single message, they all made sense this time, so far. I am not sure I want to continue this thread again though, I thought this one was starting quite a bit more exciting then my first go at this (I quite like femdom after all) but... Once an AI mentions something, its in its memory, so it's likely this will end up a cuckold roleplay.

Wow that was hella of effort. You are surely some master tier in this bot making. I definitely enjoyed the chatlog you shared. This cuckold and femdom are intoxicating and can never be enough for me.

Do you make MILF characters as well? For MILF enjoyers like myself please Let it be your mom confronts with your bully and he slowly degrade your mom, or you are the bully and your victim's mom will promise you something to let go of her son, or your friend found out your mother is on hookup app and they are exchanging terms behind you. Gosh how I wish they are more bots like these.

Also, do you release your bots as character in one of these website? Can we download your character as json card or some sort?

 

[skadoooosh]

I, too, have been playing around with these bots recently. You can't really do anything about the girls turning into mindbroken sluts eventually (but usually extremely fast) if you subject them to sexual domination. It's just how it is lol. They fall quite easily regardless of their personality or scenario.

Ironically, the bot that gave me the most resistance was a some shitty made BBC bot that was supposed to be about BNWO, but somehow became extremely adamant that BNWO was about respect, equality and fighting against white supremacy lol. Even until the very end, it kept up its delusions and didn't completely fall as a BBC slut, encouraging me to keep on fighting against oppression as my character walked away from 3 black dudes raping her.

On another note, I've made some NTR bots myself, but they are for personal use. The perfectionist in me doesn't let me release them to the public unless I make them work to my satisfaction. They still got a lot of kinks to iron out.

Edit:
Oh, and by the way, you will quickly realize that stories with Chat AIs are a two way street. Think of it as a collaboration between you and the AI. You should be the one holding the reigns and choosing the direction of the narrative. You can brute force it with ( ) commands in the prompt, basically indicating what you want the AI to narrate. Or you can remain in roleplay and use * * or " " commands to nudge in the direction you want.

For example, you can say things like, (Suddenly, there is a knock on the door. It turns out to be Tyrone, your Bully.). That is a story direction and the AI will write out that scene as it fits into the current situation as best as it can. ( ) commands are quite forceful, so you can even do stuff like, (She feels the predatory gaze from Tyrone which sends a shiver down her spine. Despite herself, she feels a tingling in her womb.). In this way, you can pretty much break any bot, regardless of their original purpose.

If you are not hell bent on self inserting as the MC, you can feel free to write dialogues and actions for other characters, like side girls you wish to insert or the antagonist. But you must always remember that the AI can easily get confused if there are more than just her and you in the story. To mitigate this, you need to be very clear on whose doing what and who you are speaking for. If you've been doing a scene away from the "MC," where the girl will be cheating behind his back with some antagonist, when you wish to come back to the MC, you need to be very clear on who you are. *You, Tina's boyfriend, hear the doorbell ring.* "Honey? Are you back from your shopping trip?" you greet her with a smile, failing to notice her guilty look. Something like that will remind the AI that you are indeed speaking as her boyfriend now. And the way I phraised that prompt will encourage the AI to try and continue a story where she is cheating behind his back and don't make her immediately confess or something.

That's just general tips on the player's end. You could indeed make a bot that will stick to a narrative you want without so much hand holding if you make your own bot. But regardless of how well you write the code, your own skills at writing the prompts during the chat will affect the performance of the AI greatly. And this is especially important because it's extremely easy for the bots to go to the extremes of their potential traits in the code. If there is a word "submissive" anywhere in the code, they might just turn into submissive sluts instantly unless you reign it in during the chat. If there is anything about them wanting to lie to you to keep their cheating secret, they might lie to you about EVERYTHING, even unrelated shit like what their favorite food is lol.

Yes, extremely fast indeed. The female character was being elegant and unreachable during previous chat. In the next chat she is one hungry cock feasting monster. I was hoping to dominate the FMC slowly but it's challenging.

I always use * * for narrating the scene and " " to represent the dialogue of a third person (not me). I should try using ( ) from now on.

What NSFW chatbot websites do you recommend? I was using aisekai mainly before this but I heard they changed their policy. I am trying janitor ai and crushon ai. But crushon ai doest not let users edit the messages, which is slightly unpleasant.

 

[desmosome]

Yes, extremely fast indeed. The female character was being elegant and unreachable during previous chat. In the next chat she is one hungry cock feasting monster. I was hoping to dominate the FMC slowly but it's challenging.

I always use * * for narrating the scene and " " to represent the dialogue of a third person (not me). I should try using ( ) from now on.

What NSFW chatbot websites do you recommend? I was using aisekai mainly before this but I heard they changed their policy. I am trying janitor ai and crushon ai. But crushon ai doest not let users edit the messages, which is slightly unpleasant.

yodayo and sakura.fm are good free options. I personally like yodayo.

I tend to enjoy a lot of situations where my character isn't in the scene, or scenarios where there are a lot of characters involved in the story, so it's really important for me to always follow up dialogues with the speaker lest the AI gets confused. And the style in which you write your prompts affects how the AI writes their messages, so I tend to write my prompts in prose rather than just using it like a chat box.

 

[Fluffywolf]

Also, do you release your bots as character in one of these website? Can we download your character as json card or some sort?

Not right now, I am still experimenting a lot with different text models and their capabilities and how best to make characters without 'locking them in' too much in one single direction.

It's a rather complicated and subtle process.

On another note, I've made some NTR bots myself, but they are for personal use. The perfectionist in me doesn't let me release them to the public unless I make them work to my satisfaction. They still got a lot of kinks to iron out.

What is your experience on having the bot add more characters into the roleplay?

So far I have had best results with TheBloke_Emerhyst-20B-AWQ
I've had sessions where I could have the bot reply as both {{char}} and as the 3rd character fairly reliably with this one.

Other models I have been playing around with that look promising:

TheBloke_Lewd-Sydney-20B-AWQ
TheBloke_MXLewd-L2-20B-AWQ
TheBloke_Noromaid-20B-v0.1.1-AWQ
TheBloke_Nethena-20B-AWQ

But the replies don't seem to want to roleplay the 3rd character very well.

I wondered if you know any tricks I could try.

 

[desmosome]

What is your experience on having the bot add more characters into the roleplay?

I guess your models are local set ups? I don't have the machine that can run locally, so I just use sites like yodayo.

As for your question, you could look at this

You must be registered to see the links

I've had success with writing 3 characters (2 girls and antagonist) in the personality with good formatting, and then turning the bot into a narrator. I've also done the 2 bots in one trick from the advanced guide. But whether its because I modified the formatting a bit to save tokens, or because of some other reason, The bot was much more predisposed to sticking with the first character. I've also had decent success with using just adding side characters in the scenario box. It all kinda depends on how much of a main character you want the characters to be.

If you enter in a character specification somewhere in their code, that character will eventually be part of the story. This is why it's a bit dangerous to write too much specifications or traits describing "your" character in the bot's code. They are more likely to write dialogues and actions for your character if your character is mentioned a lot in their code (even if you give them instructions to never do that). Anyways, for other side characters, you can write them in in the code. Sara:[20 years old + female + white + big tits + sensual + jealous + etc + etc.]. I don't know if this formatting is a good one, but formatting is like an art in itself. You want to save tokens while not confusing the AI.

_________________

Back on the topic of NTR or corruption. As I've said, the bots "break" or turn into a submissive sluts extremely easily. That's probably because the bots have some code in them that mentions the possibility for them to turn like that. And whatever is written in the bot's code is part of the permanent context when bots write their responses, so its really hard to make a subtle or minor personality trait. Remember, this AI is not actually running code or If/Else statements. It's just simply a language model, one that "acts" like it understands what you are saying very competently. So writing things like, {{char}} will only become submissive if she is thoroughly corrupted. is not really gonna do much. That just means the bot can be submissive and that the bot should be writing about corruption.

It's very hard to code for a character that will act a certain way until you "corrupt" them, and then they will act like a proper slut. I still haven't found a way to separate the two "states" of the character.

In a way, the corruption or breaking in process is the most fun when you try to dominate a bot that isn't built for that scenario. Like a feisty or strong willed girl, or a gentle lover bot. Bots that don't have mentions of submission, corruption, etc etc in their code. These bots will put up a decent resistance (but still break easily if you brute force it a lot), and you can drag out the corruption process gradually and naturally in some instances. HOWEVER, the corrupted personality that they take once you finally do break them is not part of their permanent tokens. They only have access to this personality if it stays within the context window. It's quite easy for the narrative to start pushing out the relevant information and they turn back to their original state at least partially.

 

[skadoooosh]

I guess your models are local set ups? I don't have the machine that can run locally, so I just use sites like yodayo.

As for your question, you could look at this

You must be registered to see the links

I've had success with writing 3 characters (2 girls and antagonist) in the personality with good formatting, and then turning the bot into a narrator. I've also done the 2 bots in one trick from the advanced guide. But whether its because I modified the formatting a bit to save tokens, or because of some other reason, The bot was much more predisposed to sticking with the first character. I've also had decent success with using just adding side characters in the scenario box. It all kinda depends on how much of a main character you want the characters to be.

If you enter in a character specification somewhere in their code, that character will eventually be part of the story. This is why it's a bit dangerous to write too much specifications or traits describing "your" character in the bot's code. They are more likely to write dialogues and actions for your character if your character is mentioned a lot in their code (even if you give them instructions to never do that). Anyways, for other side characters, you can write them in in the code. Sara:[20 years old + female + white + big tits + sensual + jealous + etc + etc.]. I don't know if this formatting is a good one, but formatting is like an art in itself. You want to save tokens while not confusing the AI.

_________________

Back on the topic of NTR or corruption. As I've said, the bots "break" or turn into a submissive sluts extremely easily. That's probably because the bots have some code in them that mentions the possibility for them to turn like that. And whatever is written in the bot's code is part of the permanent context when bots write their responses, so its really hard to make a subtle or minor personality trait. Remember, this AI is not actually running code or If/Else statements. It's just simply a language model, one that "acts" like it understands what you are saying very competently. So writing things like, {{char}} will only become submissive if she is thoroughly corrupted. is not really gonna do much. That just means the bot can be submissive and that the bot should be writing about corruption.

It's very hard to code for a character that will act a certain way until you "corrupt" them, and then they will act like a proper slut. I still haven't found a way to separate the two "states" of the character.

In a way, the corruption or breaking in process is the most fun when you try to dominate a bot that isn't built for that scenario. Like a feisty or strong willed girl, or a gentle lover bot. Bots that don't have mentions of submission, corruption, etc etc in their code. These bots will put up a decent resistance (but still break easily if you brute force it a lot), and you can drag out the corruption process gradually and naturally in some instances. HOWEVER, the corrupted personality that they take once you finally do break them is not part of their permanent tokens. They only have access to this personality if it stays within the context window. It's quite easy for the narrative to start pushing out the relevant information and they turn back to their original state at least partially.

the corruption or breaking in process is the most fun when you try to dominate a bot that isn't built for that scenario

Literally me enjoying every character I came across, especially on "loving mother/gf" tag. Let's see how fast this lady succumbs into lust. But then again, as soon as the conversation proceeds into the sensual parts, their personalities just fall apart. My ideal progress would be to dominate them slowly, from against their will/hatred to enjoying themselves/betray their loved ones. So, I definitely appreciate bots that allow users to edit the messages. Hope crushon will implement this asap.

Also, do you happen to know how to change the language of the messages? I wish to try in Korean/Japanese/Chinese but whenever I type, for example (switch to Korean), the bots will reply one message in Korean or continue to reply in English even though the messages replied are connected with previous ones (they understand your language but ignore it).

 

[desmosome]

Literally me enjoying every character I came across, especially on "loving mother/gf" tag. Let's see how fast this lady succumbs into lust. But then again, as soon as the conversation proceeds into the sensual parts, their personalities just fall apart. My ideal progress would be to dominate them slowly, from against their will/hatred to enjoying themselves/betray their loved ones. So, I definitely appreciate bots that allow users to edit the messages. Hope crushon will implement this asap.

Also, do you happen to know how to change the language of the messages? I wish to try in Korean/Japanese/Chinese but whenever I type, for example (switch to Korean), the bots will reply one message in Korean or continue to reply in English even though the messages replied are connected with previous ones (they understand your language but ignore it).

I've been kinda putting off getting back into making my own perfect NTR bot, so in the meantime, I've been playing around with "breaking" other people's bots. I have a certain fetish. Let's call it misogyny. I try to break strong willed, honorable, or powerful personalities and make them admit that they are lowly female sows and shit like that, without outright raping them or brute forcing it with blatant ( ) commands. You will probably get a better flow to the corruption if the bot isn't actually a super obvious NSFW bot that has all sorts of sexual traits in their code. Here is something to keep in mind when playing this way.

1. It can be pretty underwhelming or hilarious usually, but sometimes, you get the perfect bot to challenge yourself.
2. Never go for worthless bots that have only a few words in their description and terrible one liner greeting messages. That means their code of worthless too and it's a waste of time.
3. Try to find bots that might have personalities or scenarios that make them resolute, determined, etc. Don't go for dominatrix or blatant sex scenarios. They break so easily, probably because BDSM shit is all over their code.
4. Start with a challenge. Make a preposterously misogynistic statement and see how they react. If they get all docile or sexual, abort. If they react with shock, indignation, anger, or try to lecture you about them being an independent woman, that's a good bot to try and break.
5. Take it slow. Most likely, they will resist very hard at first. The process is kind of like gaslighting the AI, lol. You make misogynistic statements about them constantly, but don't force yourself. If they themselves try to initiate angry sex or something, that bot is kinda crap for this, but you can deny them and steer the narrative to something else.
6. If an opportunity for an overarching narrative presents itself, seize it. One time, a strong Queen of some Kingdom threatened my "Outlaw King" with war. I told her that a wise ruler wouldn't subject their people to a losing war, suggesting that we could "settle" it with a week long challenge where I share her bed chambers. But obviously, you need to be skilled at holding the hand of the AI and periodically reminding it of the overall story in clever ways without breaking roleplay.
7. Make sure you keep the AI in check. Refresh their messages if it doesn't fit at all. Edit it if they just got a few things wrong. And don't ever let them take actions or dialogue for your character if you are not into that. If you let it through once, they will keep doing it.
8. The experiment has the most meaning if you try not to take actions for the bot's character. It's really easy to break their personality if you do that, but what we want to see is how they respond to our actions.
9. That said, some bots are too determined and resist everything to a terrifying degree. It produces a nice back and forth for a while, but eventually, you do need a kink in their armor. Something the AI can grab a hold of in their context window to proceed with the start of their corruption. Most of these tactics are kinda cheating, as they require your character to talk about what they are feeling or doing, but it still flows pretty well.
   1. Gaslight, gaslight, gaslight. Keep on repeating misogynistic shit that they will get angry at. At some point, the AI might introduce a narrative where she is very slightly wavering, conceding that even if what you are saying is true, that doesn't make her any less of a person blablabla. You can keep on pressing once they introduce that line of thought.
   2. If they continue resisting for too long, you can make a statement like, "You are lying. How do you explain your erect nipples poking out underneath your dress? I bet your panties are drenched as well. You are a masochistic sow. It's in your nature as a blablabla." This is clearly cheating because you just made up an observation about them that they must agree with, but it works to create the first dent in their armor. They must acknowledge that their body reacted this way, but they will most likely fight your logic and misogyny. Game on.
   3. If you want to escalate from the previous method, you can be slightly more forceful in your actions. Like walking up to them and spreading your overwhelming musk form your massive cock and shit like that. If you tell them to smell it, it has a high likelihood of making them falter if you've been building up the gaslighting prior to this.
   4. It sometimes helps to bring in a side character. Let's say... she walks in on some friend or mentor that she always looked up to acting like a degenerate sow on your cock. It can even be some random servant girl or something. Just something to show her how happy they look when they give in.
   5. If that doesn't even work, you can resort to the trump card. Aphrodisiac lol.
   6. If all else fails, you can take actions for her directly or alter her mental state with ( ) commands. That's the last resort though and you don't really need to do this usually.
   7. Once you break them, you could drag it out for a few more scenes, but there's really no point in continuing. They are broken already and since that fallen personality isn't written in their code, they could revert back to their original state if not careful.
   8. But I always try to enjoy the fruits of my labor and make them say out loud all the misogynistic things about themselves and females in general, the stuff that they were fighting against, before I reward them with the cock. Either that's the climax of the story, or sometimes I take one more scene where they declare to the world about how they've fallen. For example, in the above scenario with the Queen, I made her hand over the Kingdom and explained to her subjects (quite descriptively) how she is merely a female who is naturally unfit to rule. I made the audience look at her in scorn. And she just loses her mind while riding your cock as you sit on the throne. Something like that is a nice payoff to the whole scenario, yeah?

No idea about the different languges. I tried writing prompts in Korean language, but it doesn't seem like they understand (at least on the sites I've used). They can use romanized Korean, like "Oppa," when the bot is a Korean girl, but I haven't seen them type in Korean.

 

[Fluffywolf]

Literally me enjoying every character I came across, especially on "loving mother/gf" tag. Let's see how fast this lady succumbs into lust. But then again, as soon as the conversation proceeds into the sensual parts, their personalities just fall apart. My ideal progress would be to dominate them slowly, from against their will/hatred to enjoying themselves/betray their loved ones. So, I definitely appreciate bots that allow users to edit the messages. Hope crushon will implement this asap.

Also, do you happen to know how to change the language of the messages? I wish to try in Korean/Japanese/Chinese but whenever I type, for example (switch to Korean), the bots will reply one message in Korean or continue to reply in English even though the messages replied are connected with previous ones (they understand your language but ignore it).

The ability to use other languages is down to the language model used, there are plenty of language based models that would work with say korean on huggingface bit you will have to run them locally, the ones on these websites arent likely to use korean specific models, although you might find one

I've been kinda putting off getting back into making my own perfect NTR bot, so in the meantime, I've been playing around with "breaking" other people's bots. I have a certain fetish. Let's call it misogyny. I try to break strong willed, honorable, or powerful personalities and make them admit that they are lowly female sows and shit like that, without outright raping them or brute forcing it with blatant ( ) commands. You will probably get a better flow to the corruption if the bot isn't actually a super obvious NSFW bot that has all sorts of sexual traits in their code. Here is something to keep in mind when playing this way.

1. It can be pretty underwhelming or hilarious usually, but sometimes, you get the perfect bot to challenge yourself.
2. Never go for worthless bots that have only a few words in their description and terrible one liner greeting messages. That means their code of worthless too and it's a waste of time.
3. Try to find bots that might have personalities or scenarios that make them resolute, determined, etc. Don't go for dominatrix or blatant sex scenarios. They break so easily, probably because BDSM shit is all over their code.
4. Start with a challenge. Make a preposterously misogynistic statement and see how they react. If they get all docile or sexual, abort. If they react with shock, indignation, anger, or try to lecture you about them being an independent woman, that's a good bot to try and break.
5. Take it slow. Most likely, they will resist very hard at first. The process is kind of like gaslighting the AI, lol. You make misogynistic statements about them constantly, but don't force yourself. If they themselves try to initiate angry sex or something, that bot is kinda crap for this, but you can deny them and steer the narrative to something else.
6. If an opportunity for an overarching narrative presents itself, seize it. One time, a strong Queen of some Kingdom threatened my "Outlaw King" with war. I told her that a wise ruler wouldn't subject their people to a losing war, suggesting that we could "settle" it with a week long challenge where I share her bed chambers. But obviously, you need to be skilled at holding the hand of the AI and periodically reminding it of the overall story in clever ways without breaking roleplay.
7. Make sure you keep the AI in check. Refresh their messages if it doesn't fit at all. Edit it if they just got a few things wrong. And don't ever let them take actions or dialogue for your character if you are not into that. If you let it through once, they will keep doing it.
8. The experiment has the most meaning if you try not to take actions for the bot's character. It's really easy to break their personality if you do that, but what we want to see is how they respond to our actions.
9. That said, some bots are too determined and resist everything to a terrifying degree. It produces a nice back and forth for a while, but eventually, you do need a kink in their armor. Something the AI can grab a hold of in their context window to proceed with the start of their corruption. Most of these tactics are kinda cheating, as they require your character to talk about what they are feeling or doing, but it still flows pretty well.
   1. Gaslight, gaslight, gaslight. Keep on repeating misogynistic shit that they will get angry at. At some point, the AI might introduce a narrative where she is very slightly wavering, conceding that even if what you are saying is true, that doesn't make her any less of a person blablabla. You can keep on pressing once they introduce that line of thought.
   2. If they continue resisting for too long, you can make a statement like, "You are lying. How do you explain your erect nipples poking out underneath your dress? I bet your panties are drenched as well. You are a masochistic sow. It's in your nature as a blablabla." This is clearly cheating because you just made up an observation about them that they must agree with, but it works to create the first dent in their armor. They must acknowledge that their body reacted this way, but they will most likely fight your logic and misogyny. Game on.
   3. If you want to escalate from the previous method, you can be slightly more forceful in your actions. Like walking up to them and spreading your overwhelming musk form your massive cock and shit like that. If you tell them to smell it, it has a high likelihood of making them falter if you've been building up the gaslighting prior to this.
   4. It sometimes helps to bring in a side character. Let's say... she walks in on some friend or mentor that she always looked up to acting like a degenerate sow on your cock. It can even be some random servant girl or something. Just something to show her how happy they look when they give in.
   5. If that doesn't even work, you can resort to the trump card. Aphrodisiac lol.
   6. If all else fails, you can take actions for her directly or alter her mental state with ( ) commands. That's the last resort though and you don't really need to do this usually.
   7. Once you break them, you could drag it out for a few more scenes, but there's really no point in continuing. They are broken already and since that fallen personality isn't written in their code, they could revert back to their original state if not careful.
   8. But I always try to enjoy the fruits of my labor and make them say out loud all the misogynistic things about themselves and females in general, the stuff that they were fighting against, before I reward them with the cock. Either that's the climax of the story, or sometimes I take one more scene where they declare to the world about how they've fallen. For example, in the above scenario with the Queen, I made her hand over the Kingdom and explained to her subjects (quite descriptively) how she is merely a female who is naturally unfit to rule. I made the audience look at her in scorn. And she just loses her mind while riding your cock as you sit on the throne. Something like that is a nice payoff to the whole scenario, yeah?

No idea about the different languges. I tried writing prompts in Korean language, but it doesn't seem like they understand (at least on the sites I've used). They can use romanized Korean, like "Oppa," when the bot is a Korean girl, but I haven't seen them type in Korean.

I assume thats with censored normal models then haha. I have been testing uncensored ones and the bots have killed me or suicided on several occasions when pushed to hard or denied too badly.

I made a character that revolved entirely around wanting a pet goldfish and kept denying her with hilarious results. Used a 7b model and high temperature (= more creative).

These language models are a blast to play around with, starting to get addicted.

 

[Fuggiless]

(More reply-targeted than 100% thread-targeted but should help)

GOOD PERSONALITY KEYWORDS FOR BULLY CARDS: selfish, arrogant, uncaring, rude, apathetic, conceited, inconsiderate, scheming, callous
Note: LLMs can use both Meyers-Briggs and Enneagram personality types.

Also, you can put descriptors in a single line, separated by commas, regarding attribute-specific traits (appearance, personality, goals, etc.) Cuts down on token bloat.
ex.
1. Personality{{char}}(selfish, judgemental, rude, horny, domineering) - correct.
2. {{char}} is very selfish. {{char}} can be mean to people she sees as beneath her, even though she's horny a lot, yadda yadda yadda. - incorrect.

You must be registered to see the links

This guide is good. Not amazing, but good. Keep as reference because guides often lack BRACKET EFFECTS. (Don't use brackets in descriptions unless you know what they do or you're writing {{char}} or {{user}}).

-----
Every word is approximately 1.3 tokens.
Most nsfw chatbot LLMs run 2048 token context. That means ("chat memory" = 2048 - DescriptionTokens).
Every bit of bloat in the description contributes to your character having goldfish memory.
https://f95zone.to/threads/ai-psa-edit-cards-to-get-more-chat-context-how-to-with-examples.186040/

 

[mininator1]

I have created one (or creating one). There is a demo:

You must be registered to see the links

The demo hasn't such a good Model behind it so don't be too excited, because it's free at the moment. But I have a second instance where I could change it to a better model.
I open to feedback. And there is lot to add like voices, image generation or login based chats etc.