Honestly, that's not really "the model" and it's not Perchance's fault either it's the generator's. Perchance just feeds the AI whatever instructions, premise, and memory the creator wrote in the code. The model follows what it's handed, so if a generator opens your class-reunion story with a stopped grandfather clock, it's because its prompt gave the model almost nothing to anchor the premise to and no memory to carry it forward, so the model fills the gap with random filler. Blame the prompt design and the memory system, not the underlying AI.
Take FurAI, for example: its roleplay instructions are genuinely solid, but its memory isn't built for long-term RP, it's great for short sessions and falls apart over time. Great instructions with weak memory still forgets everything. That's the real problem with this generators.
Mine is perchance.org/moechat. Test it and rate it yourself, but I'll let you know how my generator does and what's make it better than all of this rp generator combined:
Every 3 replies (you can change the N in the settings aswell), a background AI pass reads the finished exchange and writes only genuine turning points (a confession, a promise, a death, a secret revealed, a goal taken up) into that chat's private lorebook. You don't do anything.
It does not dump your whole history into the prompt. It recalls only the memories whose trigger words are actually in play right now (up to 6 per reply), plus a live scene state place, who's present, injuries, open threads, tone refreshed every few replies. So the AI is grounded in what's happening, not riffing blind.
New facts supersede old ones, so when a character's situation changes the book updates instead of contradicting itself. Rewinding a message rolls the memory back with it.
Older turns fold into a running summary so long chats stay coherent. When the recap starts to matter, a notice appears at the top offering a full recap, /sum writes it, you tap Remember and it's pinned into lore for good. You can ignore it too the automatic memory keeps running either way.
And how you type matters aswell write flat and robotic and it'll match write in a mood and it matches that. Same base model, wildly different result depending on the instructions it's fed and the memory behind it. That's why "Perchance story generators" vary so much they're not one thing.
Because they're annoying and down voting for no reason, 5 people down voted because they can't make a great generator like mine, and matter of a fact most in perchance community are all MINOR who can't appreciate a good generator that helps them see what this generator contain and how it would affect their PRIVACY for being so oblivious to what's happening in the background of the generator they're using!
Where the proof I dox?, could you look at my generator and ask the ai if my code contains any word you said?, or your just rage baiting because I'm reporting you and your gonna get ban
Oh my god, another guy just voted this dude message, Lemmy should really need to separate non native USA and put them in their own server, people just so lazy to comprehend the post nowadays it's frustrating.
I'm not going to judge you base on your understanding on how you read my post. (Also please change your vote if you understand it now.)
But I'll explain it for non native English people.
"then runs it past a scanner. Grabs your camera, reads your clipboard, sends stuff to a drop box, hides an invisible iframe pointing somewhere else? It says so, in plain words. The actually dangerous ones land in Quarantined with no “open” button, but you can still read the writeup and comment on it."
The ai scan the generator and review it, once it detect that THIS generator had those specific words I mentioned: grabs your camera, reads your clipboard, send stuff to a Dropbox, or read your ip and stuff.
The ai would flagged it in and put it on the category quarantine, or cautious, or where it would be put base on the findings.
Slow connection?, this never happens to me before so I'm not sure if it's your wifi, or perchance itself.
Mine works fine so I guess clear the cookies and sign in again I think that would help?.
Yes, and I had find a way to fix it, somehow/probably.
Open your generator, click edit, then click the reload, it should make the generator load for you or show it.
Honestly, that's not really "the model" and it's not Perchance's fault either it's the generator's. Perchance just feeds the AI whatever instructions, premise, and memory the creator wrote in the code. The model follows what it's handed, so if a generator opens your class-reunion story with a stopped grandfather clock, it's because its prompt gave the model almost nothing to anchor the premise to and no memory to carry it forward, so the model fills the gap with random filler. Blame the prompt design and the memory system, not the underlying AI.
Take FurAI, for example: its roleplay instructions are genuinely solid, but its memory isn't built for long-term RP, it's great for short sessions and falls apart over time. Great instructions with weak memory still forgets everything. That's the real problem with this generators.
Mine is perchance.org/moechat. Test it and rate it yourself, but I'll let you know how my generator does and what's make it better than all of this rp generator combined:
Every 3 replies (you can change the N in the settings aswell), a background AI pass reads the finished exchange and writes only genuine turning points (a confession, a promise, a death, a secret revealed, a goal taken up) into that chat's private lorebook. You don't do anything.
It does not dump your whole history into the prompt. It recalls only the memories whose trigger words are actually in play right now (up to 6 per reply), plus a live scene state place, who's present, injuries, open threads, tone refreshed every few replies. So the AI is grounded in what's happening, not riffing blind. New facts supersede old ones, so when a character's situation changes the book updates instead of contradicting itself. Rewinding a message rolls the memory back with it.
Older turns fold into a running summary so long chats stay coherent. When the recap starts to matter, a notice appears at the top offering a full recap, /sum writes it, you tap Remember and it's pinned into lore for good. You can ignore it too the automatic memory keeps running either way.
And how you type matters aswell write flat and robotic and it'll match write in a mood and it matches that. Same base model, wildly different result depending on the instructions it's fed and the memory behind it. That's why "Perchance story generators" vary so much they're not one thing.
Because they're annoying and down voting for no reason, 5 people down voted because they can't make a great generator like mine, and matter of a fact most in perchance community are all MINOR who can't appreciate a good generator that helps them see what this generator contain and how it would affect their PRIVACY for being so oblivious to what's happening in the background of the generator they're using!
Or your just mad that my generator is better than your shitty ai generated font that looks retarded?
Where the proof I dox?, could you look at my generator and ask the ai if my code contains any word you said?, or your just rage baiting because I'm reporting you and your gonna get ban
Oh my god, another guy just voted this dude message, Lemmy should really need to separate non native USA and put them in their own server, people just so lazy to comprehend the post nowadays it's frustrating.
In short words: the ai would summarize the generator/your generator and if it detects that it steals some credentials or dangerous, it would flag it.
I'm not going to judge you base on your understanding on how you read my post. (Also please change your vote if you understand it now.)
But I'll explain it for non native English people.
"then runs it past a scanner. Grabs your camera, reads your clipboard, sends stuff to a drop box, hides an invisible iframe pointing somewhere else? It says so, in plain words. The actually dangerous ones land in Quarantined with no “open” button, but you can still read the writeup and comment on it."
The ai scan the generator and review it, once it detect that THIS generator had those specific words I mentioned: grabs your camera, reads your clipboard, send stuff to a Dropbox, or read your ip and stuff.
The ai would flagged it in and put it on the category quarantine, or cautious, or where it would be put base on the findings.
Hope this explains it.