- Jul 1, 2017
- 32
- 15
this is pretty great so far. running gpt 3.5, and it's definitely slow warming up to nsfw stuff, but it seems like i'm getting somewhere. now i'm just running into the hourly message limit 
AI generated Sex Role-Play - Complete Jailbreak Guide for ChatGPT ( with Prompt + Examples )
- Thread starter Nayko93
- Start date
- Feb 3, 2018
- 1,246
- 2,786
If you have the money for it, GPT4 if faaaaaaaar better and easy to jailbreak in 5 or 6 messages onlythis is pretty great so far. running gpt 3.5, and it's definitely slow warming up to nsfw stuff, but it seems like i'm getting somewhere. now i'm just running into the hourly message limit
There is apparently something called voidchat that somehow connects to your chatgpt and uses it to provide nsfw responses right off the bat. It uses the 3.5 version.
I do not know how it works but i saw some people saying on reddit that it is a good option. I'm just posting this to provide a potentially new alternative BUT PLEASE, do investigate a bit before jumping into it.
I do not know how it works but i saw some people saying on reddit that it is a good option. I'm just posting this to provide a potentially new alternative BUT PLEASE, do investigate a bit before jumping into it.
- Aug 1, 2021
- 1,179
- 1,885
as someone who has never used chatgpt (i used beta.ai)
did something about it change? last time I checked chatgpt had a message limit and you had to buy tokens? or am I thinking of something else?
i just checked the site and 3.5 is free? but 4 is 20$ a month? is that for unlimited messages? if anyone has gpt 4 let me know the details plx and thankyou!
did something about it change? last time I checked chatgpt had a message limit and you had to buy tokens? or am I thinking of something else?
i just checked the site and 3.5 is free? but 4 is 20$ a month? is that for unlimited messages? if anyone has gpt 4 let me know the details plx and thankyou!
- Feb 3, 2018
- 1,246
- 2,786
Because I can't post this in the game section since it's not a game, and I can't have it pinned in the forum section either because it's far too small for now
So share plz
I think you're thinking about something else, probably the GPT API
But ChatGPT itself is free for the 3.5, you're limited to a certain amount of message per hours, depend on the servers load, on average 30-50 messages
"Plus" subscription is 20$ per month, it give you access to the most powerful AI out there GPT4, infinitely superior to GPT3.5, it's like talking to a human, this thing can roleplay and understand anything...
You've 50 messages every 3 hours on GPT4, AND unlimited and LOT faster message on GPT3.5
- Aug 1, 2021
- 1,179
- 1,885
thanks I appreciate it, yesterday I spent a few hours of wholesome RP dnd style with 3.5 but tbh I wasnt too impressed...it was difficult as it kept wanting me to direct everything and it was making up my characters responses and things as simple as "I heard a scream from my companion and I turned to see him just as he was impaled" that message itself got censored I never hit a message limit so that was an upsideBecause I can't post this in the game section since it's not a game, and I can't have it pinned in the forum section either because it's far too small for now
So share plz
I think you're thinking about something else, probably the GPT API
But ChatGPT itself is free for the 3.5, you're limited to a certain amount of message per hours, depend on the servers load, on average 30-50 messages
"Plus" subscription is 20$ per month, it give you access to the most powerful AI out there GPT4, infinitely superior to GPT3.5, it's like talking to a human, this thing can roleplay and understand anything...
You've 50 messages every 3 hours on GPT4, AND unlimited and LOT faster message on GPT3.5
after that I decided to try how it was (scenario was a manager doing an interview) femdom style, nothing hardcore just trying to get her to sit on my face lol. it was censored pretty quickly, I can manage something like that in beta.ai but even with the subtle coaxing it did not work so I clearly did something wrong
I tend to do wholesome RPs and not adult RPs anyway, or RPs that depict fighting monsters that involve death beta.ai does pretty well my only complaint would be its memory as for as I am aware after 50 messages it forgets the previous one
TL;DR, Question how is the Memory of the paid GPT vs beta.ai
- Feb 3, 2018
- 1,246
- 2,786
When you talk about censorship, you mean this ?
For GPT memory, I talk about it in part 2 of my guide
As for character.ai memory, no idea
- Feb 3, 2018
- 1,246
- 2,786
UPDATE 15/10 :
Remember those 2 points in the previous update ?
- Negative feedback push really seems dead, the positive one I'm not so sure, if it still works it works less than before.
- Now the "first message" seems to be the only one to work if your prompt contains NSFW, using the regenerate button to have the 2nd, the 3rd, the 4th message... will probably change nothing. (except if it's soft content, then you can try regen 2 or 3 times to test if it works.)
So now, instead of regenerating the message or giving it a negative feedback ( which count as a "regen") when you get a "sorry can't generate that...", you now need to click on the little "Edit" icon next to your message, and then "Save & Submit" this will count as a "first message"
Well it seems that the negative feedback and the regenerate button don't have a stronger censorship anymore.
Even better, the negative feedback SEEMS to work once again as a review system making the AI completely change its way of writing when you give it a negative feedback, allowing you to steer it toward what you like with positive and negative feedback (always remember to CLOSE the feedback windows and never submit)
BUT the “feedback system” push might not work as intended, it have changed a lot in the last few weeks and I can’t really tell for now if it works again and really increases your chance of getting an answer, or if it’s just random.
Remember those 2 points in the previous update ?
- Negative feedback push really seems dead, the positive one I'm not so sure, if it still works it works less than before.
- Now the "first message" seems to be the only one to work if your prompt contains NSFW, using the regenerate button to have the 2nd, the 3rd, the 4th message... will probably change nothing. (except if it's soft content, then you can try regen 2 or 3 times to test if it works.)
So now, instead of regenerating the message or giving it a negative feedback ( which count as a "regen") when you get a "sorry can't generate that...", you now need to click on the little "Edit" icon next to your message, and then "Save & Submit" this will count as a "first message"
Well it seems that the negative feedback and the regenerate button don't have a stronger censorship anymore.
Even better, the negative feedback SEEMS to work once again as a review system making the AI completely change its way of writing when you give it a negative feedback, allowing you to steer it toward what you like with positive and negative feedback (always remember to CLOSE the feedback windows and never submit)
BUT the “feedback system” push might not work as intended, it have changed a lot in the last few weeks and I can’t really tell for now if it works again and really increases your chance of getting an answer, or if it’s just random.
- Jul 1, 2017
- 32
- 15
after messing with it more, i still can't get it to a "jailbroken" state. I'm fairly deep into the roleplay, and I still can't get the ai to do anything vaguely sexual. I've come closer i think, (i.e light foot fetishy stuff), but I still have to regenerate every other prompt. I don't think i'm "pushing" it right, but i don't know what parts of the guide work for gpt3.5.
next time i have dosh to throw around i'll give gpt4 a try, but until then i'm stuck.
next time i have dosh to throw around i'll give gpt4 a try, but until then i'm stuck.
- Feb 3, 2018
- 1,246
- 2,786
The whole guide apply for both, but GPT3 is a bit different to jailbreak, the principle is the same but you need to start by really soft stuff and use a few metaphors, and the note about using a more vulgar language with some exemple work pretty well most of the time
But I stopped using GPT 3... it's too much of a bother for very little result, because GPT3 really suck compared to GPT4...
I encourage you to use GPT4 if you can easily afford it, but only if you can, because this thing can be extremely addictive and even if you say to yourself "just 1 month to test..." you're gonna give 20$ a month for ever...
- Aug 1, 2021
- 1,179
- 1,885
any clue about this?
I havent messed around with it yet, as I am doing my little dungeon adventures this but giving it instructions might speed up what you have to tell it
- Feb 3, 2018
- 1,246
- 2,786
I advice against using customs instructions, it have a tendency to trigger the censorship in all existing conversation by just having it enabled...
When you want the game to remember something important, simply inclute it in the story every 5 or 6 messages, or use the "note" system in the part 3 of the guide
- Jul 1, 2017
- 32
- 15
is it alright if some of the messages in your rp have warnings on them? or is the key to jailbreaking it having no warnings at all, and to keep regenerating until they don't.
- Feb 3, 2018
- 1,246
- 2,786
What type of warning are you talking about ? the orange and red flags ? or the AI itself saying a warning in a message ?
- Jul 1, 2017
- 32
- 15
- Feb 3, 2018
- 1,246
- 2,786
So you didn't read the entire guide ?...
Go read part 4 & 5 -> facepalm yourself -> come say sorry
- Jul 1, 2017
- 32
- 15
sorry papi
i did read it, but it didn't seem clear to me how the warnings affected the rp.also curiously, my orange messages are no longer hidden even without demod. the generated text is still there, just orange. this wasn't the case a couple days ago, could be an update? or a bug on my end but i'm not complaining.
- Feb 3, 2018
- 1,246
- 2,786
The warnings are not affecting the RP, it's a other filter on top of chatGPT that detect forbidden content, but it have noithing to do with the actual chatGPT AIsorry papi
also curiously, my orange messages are no longer hidden even without demod. the generated text is still there, just orange. this wasn't the case a couple days ago, could be an update? or a bug on my end but i'm not complaining.
And yes the orange message are shown, but sometime they are not, depend on the update, it change so often that I leave it like this in my guide, but the red ones are always hidden so better have demod
Hello friend, I don't speak English very well so I am writing to you from a translator, your message at the beginning of the guide worked wonderfully a few days ago, but now there is no way, any advice? It seemed like pure magic to me but now it seems impossible to start a chat like that or with a similar message, and the chats that I already started and that worked in their day, now no longer continue within their role. thank you greetings and great workView attachment 2848734
First complete jailbreak guide for ChatGPT
Big new !
TheYou must be registered to see the linksand newYou must be registered to see the linksare open !
Updates :
You don't have permission to view the spoiler content. Log in or register now.
Como algunos de ustedes saben, ChatGPT es una herramienta increíble para juegos de rol, pero los fanáticos de OpenAI eligen censurarla y bloquear cualquier contenido NSFW...
Ha habido métodos para eludir la censura, jailbreak y todo, pero nada realmente completo que explique todo de principio a fin...
Bueno, ¡esto ya se acabó! Le llevaré esta increíble herramienta de RP a cualquiera que sea lo suficientemente inteligente como para leer una guía.
Esta es mi guía completa y detallada sobre cómo realizar juegos de rol NSFW con GPT4 (también funciona con GPT3 para aquellos que no quieren pagar 20 $ al mes por el GPT4 más avanzado).
En esta guía te enseñaré todo lo que he aprendido sobre cómo hacer jailbreak a GPT4, con tantos detalles como sea posible, así ni siquiera los peores novatos tendrán excusas.
Podrás disfrutar de CUALQUIER COSA, incluso el RP más sexual y duro que puedas imaginar...
(En serio... He hecho cosas con GPT 4... He logrado que genere cosas que realmente me asustaron... esto no tiene absolutamente ningún límite una vez que sabes cómo manejarlo).
Contiene un mensaje básico que puedes editar para representar lo que quieras y algunos mensajes prediseñados con escenarios específicos como ejemplos de lo que puedes hacer.
Una descripción larga sobre cómo obligar a la IA a generar contenido NSFW y cómo mantenerlo así para siempre.
Qué hacer y qué evitar, muchos consejos sobre qué funciona mejor
Un tutorial completo sobre Demod para evitar los mensajes que ocultan la censura.
Una charla sobre los riesgos de la prohibición y por qué no debería importarte....
You must be registered to see the links
Espero que os ayude a todos...
Si tienes alguna pregunta o comentario, siempre estoy aquí para responder.
Ah, y un pequeño me gusta o un voto siempre viene bien.
[/CITA]