I mean, when I explicitly instructed it to curse and be crass on AI Chat and Roleplay, it actually did that. But it doesn’t do any of this on its own. Sometimes it even seemed to go into pretty dark territory. It literally varies by the day. Also, if they’re actively pursuing censorship, then I don’t think it’ll do their business real good, as it literally removes the one thing Perchance had over all other AI generators. We’re already seeing long time users leaving. Also, I’m not too familiar with model pricing, but are Google’s models cheaper than DeepSeek’s. If so, then I truly believe THAT’S the reason they’re using it.

I don’t really think it’s censorship on their part. I think it’s censorship hardcoded into the LLM itself. This model appears to be a Google model, and we all know how petty censored Google’s models are. If it were censorship on their part, I’m pretty sure it would’ve outright refused to generate anything. This just rushes through them, which to me is a telltale sign of passive LLM censorship when it doesn’t have the ability to outright refuse. Also, another to note, is that it isn’t just smut that’s been censored. Basic swear words have also been censored, which is even further proof that this is in-LLM censoring and not censorship from Perchance’s part.

It wasn’t just a model switch; it was a complete fuckup.

What on earth has happened here? No one knows.

Slang? Gone. Cursing? Let’s keep it PG-13, while wondering about the logistics of what possessed the dev to implement this godawful model.

No, but on a more serious note, I actually liked this model at first… not anymore though. Seriously, fucking c.ai provides a better experience than this now. I’ve practically abandoned the platform now and moved to this instead: solania.ai. It’s limited, but it uses the old LLM, so that immediately makes it better than Perchance atm.

Until the dev removes this model, Perchance is dead to me.

Here are my complaints tho Bad things

• As you mentioned, the dialogue, while better in most areas, does come off as a little stilted. It’ll not use contractions where they’d usually be and is very conservative with its use of swear words, which adds to the artificiality.
• The “,you X” tags are starting to irk me. This is a trait found in other ai chatbots like c.ai, which gets really annoying after a while.
• The worst one by far: It’s not X (SEMI-COLON) it’s Y. Bafflingly this was the ONLY and I mean ONLY bottom-of-the-barrel AI trope the previous model didn’t use regularly, but here it is, in all its pretentious glory. The use of a semi-colon instead of a comma just comes off like a uni student trying to appear lofty on an essay.
• The issue that probably fuels most of the hate for this model: It isn’t as emotional. This is kinda a 50/50 for me.
• It doesn’t hallucinate extra info on a character/place etc. Even tho the old model did this pretty poorly and stereotypically most of the time, it at least added some convenience. Good things
• No more ‘like’ or ‘than’ statements in every sentence.
• No more hanging on one city/object/person etc. Always took a stereotypically related object/event from the theme you were writing about and hanged on it ad infinitum.
• As you said: No more obsession with smells, which nearly drove me to offing myself.

Also, I’d recommend not using the regular story generator and using the enhanced version instead. It doesn’t have as much customizability but it works perfectly on my side. I also mostly use the AI Chat and Roleplay site as I usually wanna emulate conversations, not stories.

It’s a shame posts praising the new model are so vehemently downvoted (Guess people really like similes), but that’s a given with any model change. The previous two models drove me to drink, but this one has really been a joy. Sure, it has its issues (It’s not X, it’s Y: repetition issues etc.). The previous model was execrably repetitive, reliant on similes, overly metaphorical and descriptive, and extremely stubborn. It was really a pain in the ass to use.

Good to hear someone else shares my opinion.

To be honest, it comes down to personal preference. I mean, it absolutely has its issues (It’s not X (SEMI-COLON) it’s Y, excessive repetition etc.), but I’d still take it anytime over both the “Elena CasteLLamos” or “Deep like a Seek”.

Ah yes. I still remember the good ol’ days of Elena Castellanos.

I’d have to completely disagree. The new model has been a joy to work with so far.

• Gone are the constant flowery descriptions of useless environmental objects. (Especially its hardcoded obsessions with smell)
• Gone is the constant hanging on one topic/object etc.
• Gone is the pretentious, twee, stereotypical dialogue, which was execrably dependant on similes.
• Gone is the stubborn refusal to follow instructions on how to write. (It actually does what it’s told. I asked it to write only 2-3 line of dialogue per character and that’s exactly what it did.)
• The auto-improve feature actually works now.
• This AI can actually think for itself and doesn’t need constant nudging and correcting.

I understand you might be a bit distraught over this, but think about why the AI turned ‘robotic’: because people were sick of the constant flowery, overly verbose, filler descriptions of useless or even useful items.

I’m very glad about these changes, but I do think a post should’ve been made just to alert everyone. A rollback I believe is impossible, since dev probably spent months on this and is definitely not gonna abandon it now.

I’ve been getting this problem too. Characters that are supposed to be neutral to my character’s preference act all gaslighting and narcissistic, constantly negating my statements with curt words like ‘Bullshit’. The AI seems to follow what IT please, not what you please. It’s infuriating.

I have to agree here. The new model is only marginally better in some areas. Yet it still suffers from:

1. Overly metaphorical descriptions of useless/non-plot driving environmental objects and phenomena. Still seems to have a hardcoded obsession with smells (Even in dialogue)
2. Obsession with stage directions. (Knuckles whiten etc.)
3. Still has a limited pool of names, referentials, examples.

I also found quite a few problems with dialogue generation

1. Tweeness and Sentimentality: “Remember when Grandma (Always referred to as a proper noun regardless of relationship to speaker) used to bring us freshly baked cookies to school?”
2. Gaslighting and Negation: I don’t know if this is a personal problem, but the characters seem to be extremely gaslighting (I didn’t tell them to be that) and constantly negate my statements with ‘Bullshit’. Gets very agitating after a while.
3. Tries way to hard to be funny and falls completely flat: “This place smells like regret and stale beer.” (Two problems in one: obsession with smells and an overused AI trope) Other than that, the dialogue is slightly better and less all-out than the previous model.

Outside knowledge is satisfactory but not fantastic.

So yeah, it’s better and perfectly fine for a free tool, but not nearly up to ChatGPT levels.