AI sites and tricks

[TT2000]

With OpenAI slamming down the banhammer the last few days, improper use of Caveduck's OpenAI might be coming to an end. Assume your days are numbered, enjoy while it lasts.

The profecy was true...

damn

[DrakeX]

But its just filtering explicit nsfw content or vore of any kind?. Because even on the more sfw-ish versions of vore caveduck has always gave me much better answers than for example c.ai

Edit: btw. Caveduck isn't taking points from me today no matter how much i chat for some reason

Mandarin (caveduck's own AI) is completely free and the default. Then there's GPT-3 for 6 points a reply and GPT-4 for 120 points a reply. You're likely using Mandarin. Check the bottom-right.

Yeah you were right. I noticed the answers i was getting were not as detailed as before.

I still wonder if the new filter will involve only the most explicit content (cv, hv, Etc). Or if more tame content (soft vore, non graphic digestion, absorbtion etc) will also be included in this

[MerrySue1982]

AI already seems to tend towards fetish sometimes. I not only player around with diapers. I also played around with bikini and ovens. And the Ai keeps breaking it's own rules with the bikini it comes up and i explicitly have to tell him it's allowed to be sexy but keep it SFW^^

[AlluringPredation]

I laugh at your points with my unlimited NovelAI generations.

[TT2000]

Well, given current events, I decided to port over Nancy over to charstar. I'd like to see how she fares up with other models. Though one thing's certain, she's still unable to die!

Nancy (charstar):
https://charstar.ai/char/e9d48afa-2f31- ... 6249d1fe34

[hornidood1]

The filters are on now and you need an API key for unfiltered GPT-3 or for GPT-4 at all. Safe GPT-3 is basically Character AI with HearMyWaifu active.

Mandarin is still free and unfiltered, but it is... not good. If we're willing to stick with it in its rough days, maybe they will be able to improve it and we can have essentially a competently run Character AI, but we also may not. For what its worth, Mandarin is trying the best it can.

[Alphapred96]

I'm guessing there's no easy way to get API keys, eh?

[BladeWard]

I'm guessing there's no easy way to get API keys, eh?

Yyyep, since you gotta pay for the AI to even get the API key.

[DrakeX]

Wow it's definitely dead. Just the smallest amount of violence will trigger the filters. Let alone mention anything vore related regardless if it's sfw or not. I tried replicating a previous session with Mandarin but its baad. I get better responses with c.AI

Atleast it doesn't delete previous dialogues so if you have your previous rp sessions saved they will still be there

[Loker]

Regarding Charstar and Spicychat, they would be usable but why the hell can't they have password login and must send verification to email? I very much prefer not to have my mail full of that stuff.

[hernextmeal]

Regarding Charstar and Spicychat, they would be usable but why the hell can't they have password login and must send verification to email? I very much prefer not to have my mail full of that stuff.

Because passwords are basically useless in terms of security these days, and yet they don't want to force SMS verification like OpenAI does.

They could use TOTP (aka Google Authenticator/Authy/Microsoft Authenticator/etc...), but most users aren't going to install authenticator apps so a lot of sites don't bother with them.

[Thadeu]

Gotta love caveduck's reasoning. ''Oh we have to update and censor 99% of our bots because of gpt rules''.... well, the rules didn't get made now did they? They created this website, with freedom, no censorship and after most of their website became smut, they were like ''oh btw, we are taking it all away cause of the rules'', why make it nsfw in the first place then?

[AlluringPredation]

Gotta love caveduck's reasoning. ''Oh we have to update and censor 99% of our bots because of gpt rules''.... well, the rules didn't get made now did they? They created this website, with freedom, no censorship and after most of their website became smut, they were like ''oh btw, we are taking it all away cause of the rules'', why make it nsfw in the first place then?

Because a lot of these companies don't pay attention to OpenAI's TOS, and think they've stumbled onto an easy cash cow. Make a website, create a front-end for OpenAI, take money.
Opposed to fine-tuning a Llama2-13b model that would underperform NovelAI's proprietary model. Or Llama2-70b model for a nice user experience but requiring preparation and operation costs. Unfortunately no middle ground 30b range model since it got flubbed somehow by Meta. Maybe by Llama3 we'll have companies stop trying to take the easy street, but my guess is No.

[RoTheHuman]

Sigh, I'm glad I enjoyed caveduck while it lasted! Hopefully something else will come along soon.

I wish it were possible to self-host a decent RP-capable chatbot, but all of the ones I've tried were utterly garbage. Granted I don't have a system with an absolutely massive GPU and I'm not willing to use Google's online stuff since that's a privacy nightmare so I'm limited in what I can run.

[Bright]

How much better was the 4 version compared to the 3.5 version?

[Silversepiroth]

How much better was the 4 version compared to the 3.5 version?

The 3.5 version was great for most of the time, but there were moments where I had to bite the bullet & use 4 to get the generation to go in the correct direction on its own. If nothing else, it handles innuendo & nuance a lot better than 3.5 does

[OCGTSHero]

whelp caveduck is dead for NSFW... now

[Tira]

whelp caveduck is dead for NSFW... now

Oh well, seems like it's time to... search for yet another site without filters!

[dragorise20]

spicy chat ai still works but it's hard to make vore with it

[ANW]

whelp caveduck is dead for NSFW... now

Not really. I had a good time with a young sister that wants her older brother (You) to eat her.
I might not be good with vore, but the free mandarin choice did good with sex.
So if you want NSFW, you might have to do a little editing or persuasion, but it's still viable.