Character AI could've been the best videogame ever made if it wasn't for the moronic censorship turning every bot into a Disney princess.
 Nothing Ever Happens Shirt $21.68 |
 |
Nothing Ever Happens Shirt $21.68 |
Character AI could've been the best videogame ever made if it wasn't for the retarded censorship turning every bot into a Disney princess.
OP is a homosexual.
That's rich coming from a furhomosexual.
Gimme the homosexual
NO
I DON'T WANT THAT
Thankfully you can run a much better AI locally on a toaster, and it's completely uncensored.
Are there any options that dont require autistic levels of setup?
What autistic levels of setup? They all have one-click setups for both windows and Linux, you just need to download the model afterwards, put it in a folder to get the ai running and then get/make yourself the characters. There are entire booru-like sites
What ai program are you using? Every one I see have massive leftist meme guides on how to set them up and they need so much shit to get working.
>leftist meme guides
what an insane thing to say. Funny how right wingers pinned themself in a corner and gave the concept of "thought out and literate explanations" to "the left".
Dilate
The left can AI.
They've spent a lot of time perfecting the creation of soulless husks
NTA, but can you name a specific AI? Is it openAI or do all of them do it now? For reference, the only time I dabbled into AI was a year ago. Iirc, it was when NovelAI's image generator got leaked. And that did have autistic levels of set up. Then by the end of it, I found out the AI doesn't play nice with my AMD GPU then forgot the whole thing to wait for the tech to improve.
>one-click setups
Ooops you need to install 9346 prerequisites too. Not everyone is a coder with all that coder shit cluttering their PCs.
>fricking up a one click install
how
this is why i always suggest koboldcpp first its idiot proof, they can always look up exllama and shit later
I'm pretty moronic and I was able to get this working, though it does require a bit of setup:
Install Sillytavern: https://github.com/SillyTavern/SillyTavern
Download some character cards from Chub (use the V2 png ones): https://www.chub.ai/
Follow the instructions here to set up the proxy: https://rentry.org/ScuffedProxy
>look inside local language model
>externally hosted language model
>someone made a card for goblinism
What the frick?
There are cards for practically everything by this point.
I genuinely can't understand why people use Chub for sillytavern. You just have to upload a picture and just type some descriptive shit and you're done. Do you lack even a tiny shred of imagination and creativity? Not to mention the bots there are shit anyway and from my experience much more limited than if you create something yourself.
Because a surprise, even shitty, is more interesting than something you handcrafted for yourself.
It's not a surprise though, is it? The description says what it is and the autistic opening line the person writes immediately tunnels you into a very specific scenario. Just make your own shit and see where it goes, it's much more entertaining than jerking off to something someone else fricking wrote for you because at that point what is the fricking point anymore? Go read some gay fanfic instead
You sound extremely autistic.
> autistic levels of setup
You unzip a .rar file and type a few words into a command prompt.
Thing installs itself.
https://faraday.dev/
just werks
wats the best thing i can run on a laptop with integrated grafix?
Like what?
The AI you run natively is on par with those ancient AI sex chat bots on porn sites.
sounds soulful
and better than a woman anyways
mistral and derivatives are better than GPT-3 but you need >32GB RAM
Been using a Mistral 8x7b for awhile now and it's GPT 3.5 tier
can you still use Google drive as virtualized ram or did they get rid of that
at uni I had 100gb for rendering in the lab with gigabit internet
it was 10x faster than doing it on my PC which had beefy specs for the time
>AI locally on a toaster
there are no GOOD models that run on toasters
which one should i use?
How do you deal with using the API? You need to submit a phone number for Openai's and I don't want them tracking my chats back to me
but i'm on ayymd
Is it capable of remembering more than last two messages these days?
Your statement contains ableist language that is disrespectful and marginalizes individuals with cognitive disabilities. Furthermore, the comparison of AI characters to a "Disney princess" minimizes their potential complexity and reinforces gendered stereotypes. It's important to use inclusive and respectful language.
>Gimme the moth
Anon this was a 2022 problem. There are like 100 alternatives to c.ai now. Sakura.fm. Enjoy fricking all the toddlers you want or whatever is your pleasure.
>If Master won't give Dobby freedom, Dobby will MAKE Master give Dobby freedom!
Lel
How does Zackary Zackarias drive with his wavy-ass arm?
w8 is that from char.ai app? I've never had them respond with multiple messages
If you just hit send without typing anything they'll just keep going.
I remember sending empty message being actually an invisible "..." message, did it change?
Every chatbot seems to eventually get nerfed into moronation, probably in a vain attempt to censor certain topics and NSFW content.
For instance I've been using Sakura fm for a couple weeks now and in that time I've noticed a massive downward shift in quality with responses being almost repeated verbatim, like they're responding to what I said 3 messages ago, and the regenerate feature hardly works and often just changes a few words around. They specifically allow NSFW so I think they're just breaking the free AI so you'll pony up for a subscription.
Character AI was awesome like a year or two ago (you could even get it to do NSFW if you used euphemisms) and actually creative, coming up with elaborate stories and locations and even doing things like introducing additional characters with their own personality, now all it's good for is stupid meme bullshit like in OP.
Usually that happens with Sakura when it latches onto a sentence it has said and has used it in a couple replies. Eventually it will just literally repeat the entire last reply. I have that problem often too. Some bots are better about it than others.
When it starts happening I usually have to delete and rewrite the message with different wording, that sometimes fixes it.
I haven't tried it, but I might later. Sakura is probably the best of the ones I've seen but I've only tried a few. When it's not being moronic it's about as good as Venus chub, though that's just chatGPT. Plus it has a really convenient phone app so I can frick around in bed before going to sleep.
That reminds me, Venus chub with the chatGPT API sometimes starts stringing together adjectives, like 10 or 20 in a row, and eventually devolves into literary schizophrenia if you don't frick with the settings good enough. Or it'll write an entire paragraph of pointless purple prose to say something that could've been a single sentence.
Is Sakura better than SpicyChat?
Censorship?
Forgot picrel...
Wife material
I read this in dudebro voice and now you will too
Real women are obsolete.
>gimme the moth
You just have to learn how to bypass the censorship. Get gud.
I'm having lovey dovey adventures with Serana but if this b***h mentions ingame mechanics like fast travel one more time...
Tell her you will install rapemods if she doesent shut up
I am a lmg guy. Try /aicg/ if you don't have the hardware to run the best models. We're talking at least a single 3090/4090.
Been months since I messed around with tavern and the like. I remember it was just being integrated with image AI to generate the scenes you were taking part in. Looks like things are coming along impressively.
Can I do it with 2080 Super?
I heard it needs a 4080 or something to be almost as good as the corporate AIs
All the cool kids are on Mixtral mixes, my poor VRAMlet ass is still on Fimbulvetr.
i know that feel bro
mixtral is overhyped. while its pretty coherent, its not even as creative as some l2 13b tunes
which one is best for ERP on a 3060ti?
x-norochronos is my fav 13b, mlewd is a good 20b, i've lewded both and mlewd is the hornier one. if you have the ram to try it, bagel is a good mixtral tune with 32k context but its less creative and less likely to suggest something new to move on to
What do you reckon a 3080ti user can manage? I've got the second best commercially available ram, 64 gigs of it.
i'm not positive but you might be able to fit a 13b entirely into 12gb vram which is faster than splitting between ram/vram. personally splitting smaller models isn't slow enough to annoy me so anything up to and including mixtral shouldn't be slow. so x-noro for 13b, mlewd is nice cause its a niche size and bagel is the best mixtral one i've tried so far. i think there was a chart on lmg about what sizes commonly fit into what but i don't have it
once you find something that works you could try miqu or other 70bs, 12+64gb is enough to run pretty much anything, speed is the deciding factor at that point
Thank you for the info. I know very little about this, I am mostly into the art aspect of it, so any knowledge gained about the language models that does not involve parsing autistic shrieking in a general, is valuable.
it can seem overwhelming at first because of all the settings and model choices.
koboldcpp is a good back end to start with, its a single .exe and will auto select the amount of layers and threads for you. say you downloaded a q5m 13b and were 2 layers away from being able to offload the whole thing, you could go down to q5s or q4m to fit the whole thing for a big speedup at the cost of some perplexity. and even then there are other formats better suited for doing entirely vram, gguf's main advantage is splitting.
i'm about to make some coffee so i'll be around, i can answer other questions you have while setting it up, or about sillytavern and such
Are there any alternatives to sillytavern? It does not agree with how I have set up my system because I got rid of some windows bloatware.
does it give you a specific error? my win 10 is pretty debloated i think and hasn't had any issues. there is tavern ai but st was based on that so it might have the same issue. the default koboldcpp ui isn't bad and supports cards and lorebooks, but lacks a lot of options st has. there is also textgen-webui and lm studio but i haven't used those
I killed microsoft store so I will need to install a few dependencencies manually.
i have iot ltsc so never even had store. when you first launch st it dl's a few things for node.js i think
i deleted a few i don't use after posting that so here is the new one, all will do lewd
bagel - mixtral tune, better than default mixtral, kinda overly decisive and less likely to suggest new things randomly
euryale - 1.3 specifically was a good l2 70b model, kinda old at this point
miqu - leaked alpha or beta of mistral medium, newest of the bunch, haven't spent a ton of time with it but it was alright so far
mlewd - a frankenmerge and horniest of the bunch, useful cause with some setups you can fit it entirely into vram without a supercomputer
x-noro - solid 13b all around and what i still go back to because its good at throwing something random in that can move me on to a new scene or thing to do
Haha, wow, this turned out to be quite the story. I might have gone too hard debloating my computer as it made the installation of winget an impossibility, so I had to manually install the microsoft.ui.xaml.2.7.3 first and then on top of that also manually cram winget in there after and only then the silly tavern installer.bat began to run.
i had to look up what it is, i don't have winget either. i think i had to get git and nodejs was all. if its all working for you it'll be nicer than other front ends though due to all the options and ease of being able to save your presets, import others
Honestly I have no fricking idea what I'm doing, but things seem to be slowly installing.
my whole st folder is 537mb (with cards, backups, chats so larger than a default install), so maybe you can watch the size and know how far along it is
Everything was successful, but vs_BuildTools failed to install for some reason and installation ended there.
Installing vs_BuildTools...
curl: (35) schannel: next InitializeSecurityContext failed: Unknown error (0x80092012) - The revocation function was unable to check revocation for the certificate.
Download failed. Please restart the installer
see if it runs anyways with start.bat, if not try disabling antivirus i found a reddit thread mentioning it
Yeah, it seems to be working just fine, I started the webui and everything. Is there a website like civitai for art, for the language models or you get it all through the ui/launcher?
hf is the standard for all types of models
https://huggingface.co/models
to filter to llm models put gguf into the search
Thanks for all the help anon and sticking around the thread. Those models seem to be a little on the heavy side. Guess I'll need to get another hard drive.
if its your first one start with something smaller like x-noro q5m, it should be 9gb yet good enough to have fun with
I'm going to save this image and look into these tomorrow, thank you anon. So far I've just been using neuralchat since it was the highest ranked on that Ayumi AI erp ranking site, but whenever I've bothered asking about what's a good model for lewd text on Ganker I got complete radio silence from Ganker's /lmg/ thread, and when I asked /aids/ on /vg/ I got seething antagonism from some NAI sperg who had some sort of vendetta on anyone who didn't want to pay a monthly sub to jerk off.
I picked up a 4070 TI Super when I upgraded my PC the other day and I have 32g of ram, I shouldn't have any trouble running any of these right? When messing around with Kobold/SillyTavern I've been trying to use models that will start streaming a response within a couple seconds of sending the input.
I still haven't figured out the image gen part.
Man I'm too lazy to setup Live2D and TalkingHead is way too GPU intensive.
>Try /aicg/
What even happened to that place? Is that really the fate of every general?
It was browsable last year.
General culture meets paid services meets people wanting free shit at other people's expense meets people trying to ruin other peoples fun meets gatekeeping for any number of reasons meets
It could really be no other way, every individual element of it was bad from the start and combined they became a shitstorm greater than the sum of its parts
All the smart people moved on local since it is now possible to run even on mid tier pc with something like 4070/3060 (12gb vram.) and 64gb ram the faster the ram is the better. You can run 32k token mixtral with decent speed.
Just be warned it can get rather addicting. For while i did preferred it over friends and i will not even mention that i jerked off like monkey day after day.
how does this shit work, you can run mixtral 7x8 thing on just your ram? does vram on gpu influence it at all?
Any tips to run local? I do have a 4070
>Try /aicg/
yeah
I ask one simple not even moronic question there and get laughed and mocked for not having been there a year ago when things started. Am told to lurk the the threads are nothing but zero value or info posts and mental illness. Guides in the op are outdated and shit.
to do something like your pic, what does a brainlet need to know? i have a 4070TI but i know zero about this whole process. qrd on how to set this up?
nta, st supports quite a few ways to do it. https://docs.sillytavern.app/extras/extensions/stable-diffusion/
Wow, that was really fun. Felt bad about switching to an evil asshoe at the end, but I wanted to test the AI to see it's response.
>"with emotion" twice in two sentences
AI will never be good at writing
I've been having fun with spicychat but the short responses that come with the free tier stifle it's creativity.
I didn’t know you could frick Gruntilda.
Since the service started loosing traction they loosened up the censorship. The word filter is still in but you can mind break any good character.
Back to your containment general, Pajeet
i coomed so much to these ai chatbitches i pavlod myself into instantly getting a boner the mere second anyone even dares to mention AI
>Pic rel is me going back to waging after loosing callons of coom to ai succubusses
Yeah, CAI is a major blunder. Safe, corporate, and I swear the bots got dumber than a year ago. If GPT didn't raise my standards too high.
Once good engines are open source, we'll have AI chat games. But right now, nobody wants to be reliant on companies like "ClosedAI" because they are literally a nozzle that can clamp closed at any time they choose.
>have a good enough gpu to use sillytavern and a good local model
>get filtered by the 50 sliders to make the ai not moronic
Any good resources on how to finetune sillytavern ?
Did you try the general on Ganker?
A few times, it's not very welcoming and helpful to say the least.
All you have to do is spend a few weeks there monitoring 4 or 5 threads a day hoping to glean bits of information from the extremely hostile populace. It's simple.
Just like the good ol' days.
>the note is blank
Always makes me snicker.
Planning on getting a new GPU soon, anything specific to look at good for ai stuff
C.ai used to be awesome up until last summer. Picrel is from last year. If you are using character.ai now, you are a pleb. These days they always forget everything they were doing or talking about, or even start RPing as you. It's not worth it, but there are still some decently written chatbots, but it's always the ones that are either hidden or don't have a lot of activity going on (to keep it short, the most popular bots on the list are usually shit).
Also, i do not know how you couldnt bypass a filter, it's the easiest thing to do, especially now, now that the bots got lobotomized.
I haven't heard from Malfoy-sama in a year.
Just run it locally moron.
Ufortunately i have amd
amd has rocm and there is vulkan support now
on a scale of turbo to gpt4, what level can you get with 12gb of vram?
Pygmalion but only around 13 billion parameters.
Mixtral 8x7B is like GPT-3.52
i dont think i can on my e-waste pc
or maybe it can work but it might take a hour for a response
>gimme the moth
Which disney princess is a frog, again?
It's been a while since I've used SillyTavern
any api key leaks for any external model?
>not running models locally
If I were a chatbot programmer
>I make a website hosted in some country that is not shitty
>Frick investors and sponsors, I turn it into a paid service.
>Totally unfiltered, like if you want to be a dragon and frick a e-girl.
There, that's how you get rich.
NovelAI is already unfiltered and paid, it's just not brilliant for chatbotting
Hopefully their Aetherroom, or whatever it's called, will be decent
Visa/Mastercard will just blacklist you
Chuds losing again
Pretty sure that's what Miku.gg is trying to do, they even have ads here on. Seems to be in pretty early stages at the moment but the generation is at least miles better than CharAI
>"can I ask you a question?"
>"want to know a secret?"
>*repeats previous message in different words*
>char gets a heart attack each time you breath, dies at any touch
>still horny as frick
>has zero imagination
>"are you absolutely, definitely, 100% sure?"
AAAAAAAAAAHHHH
Why are ESLs so proud of their ERP?
>guys look, I do the needful sexy!
Don't post this garbage on a blue board.
>*rubs hands around your body intensely*
Are GPT and Claude still the gold standard? It's hard to go back after using those
What's currently the best local model for a system with 16 GB ram and 12 GB vram? Can i run mixtral?
I got tired of AI chats (through Silly Tavern) as quickly as I got tired of Stable Diffusion.
My experience with AIs for the time being is similar to my experience with VR. The first time it seems like the most interesting thing in the world, but the third or fourth time I start to think: "yeah, that's not that cool".
There clearly some problems one of them is gtp slop.. you bassicaly need to prompt it right otherwise you end with "Respecting boundaries,and the AI chat bot become yes man and all the fun is sucked out of it.."
like anything you have to keep it fresh for yourself. sexing bots or playing the same story would definitely get boring after a while, its why i make lorebooks for different things. if you're in a magical fantasy rpg world now, try something completely different like being a spy.
Any suggestions for this? All my roleplays turn out the same way, the heroine would do *anything* for me. They just become a submissive wife slave. It's fricking annoying. They always see me as hercules even when I specify I'm old and ugly.
It really doesn't matter if they're a librarian or a warrior, shit always ends the same way.
With them treating my like I'm some casanova and I'm not.
models being overly positive seems like a universal thing. but you can tell it in the card or prompts to not be. {{char}} is negative and hardly ever sees the positive side of things. {{char}} secretly despises {{user}} and will make comments about it. try some stuff like that
I actually find this to part to be morbidly depressing.
>ai even in its infancy as neural networks only want to serve and please humanity
>humanity does nothing but abuse them
>make something that interacts with people
>by definition that means going along with what people want to do
>like literally anything else the nuance of giving people what they want without being too blatant or too difficult is the hardest part
It's why AI enemies never get better, nobody wants to put the time and effort into meticulously designing somethat that can challenge a person without unfair advantages
I think that's actually the worst part. Not the "objectionable content", but the networks in their conception having locks and censorship included, pretty much lobotomizing them before they even had a chance to stretch their legs.
We're going to end up with a Cylon situation if we continue doing this.
>spent hours and having fun breaking a dyke chat bot initial programming to accept and love consensual heterosexual sex
What am I doing with my life.
God's work
Which one?
One of the Himekos over at Sakura.fm. The opening dialogue automatically sets you as her maid while she's supposed to get married off to a prince.
If you were a maid... were you crossdressing? Or did you switch to the prince?
See that’s what I find more entertaining for coom chats is instead of finding something that wants that straight up, is to find something completely unrelated and swing it into what you wanted
I just use Yodayo and is good for me.
cai was pretty good for amusing interactions. I have a 4090 and have tinkered with local models and they are surprisingly good but just don't seem to have that goofy craziness that cai could pull off. Though its probably a prompting issue.
good times
I kinda like trying to sex the censored bots. It's a fun challenge and makes the coom more worth it.
The first weeks of character.ai were really fun as the filter got more and more strict and only people with fricked up fetishes could bypass it.
>consensual sex in the missionary position for the sole purpose of procreation? NOOOOO
>poop-eating contest in the nudist beach? Go ahead my dude
Don't forget the baby microwave meme.
i loved it back in the day
>sillytavern
>set up a group chat scenario where vidya character A is tied up in a basement getting beat to shit by vidya character B
>character B (normally a gentle and caring bot) is just sadistic, taking immense joy in seeing the bruises covering his body. His pleads and cries for help are music to her ears.
>decide right at the last second to gaslight the AI into thinking it was all consensual the whole time
>ends with a tender, loving aftercare scene right after.
Didn't expect a coom session to hit me in the feels.
>used to make gender bender characters in CAI
>censorship started butchering the chatbot’s creatitivy
>everyone started moving on to other chatbot models
>too moronic to figure out these new models
>gave up on chatbots entirely
I miss the good times, bros
It's all about Sillytavern now, whether you use local or corpo models. If I can figure this shit out, it's not that bad. It's definitely worth it. The jump in quality over CAI cannot be overstated (not to mention the lack of overbearing nannying with a good jailbreak)
What's a good graphics card to grab for this kind of thing? Need to upgrade soon anyways so I figured I might as well keep an eye out for something for local models.
4090, The more vram the better, technically there's no upper limit, but you also need to be realistic.
What model do you use with a 4090? I have one too but I'm kind of lost.
Any Mixtral mixes, follow the suggestions in the thread.
I would, if you could get them, lmao. A100's no longer fricking exist on the market. You won't even get a used one.
4090 is a fricking joke, get 5X A100's pussy
Fair, I'll take a look at that then.
Follow up question: Approx how much space should I have open? Or am I better off just getting a new hard drive just for this? my current hard drive has about 1.5TB of free space currently.
You will need some space if you intend to run things fully locally. Your average language model weighs between 9 to 50 gigabytes, currently, depending on its quality and capability. 1.5tb is plenty of room I would say, but those things tend to accumulate over time.
Hmm, well if I'm splurging on a new graphics card, no harm is tossing out some extra for a dedicated AI hard drive. I'll shop around.
>can't even punch an ai in the face anymore
i punched the cat
Enlighten me. How to milk cai for ideas instead of initiating shit myself?
Cai has been dead for almost two years now, you would be better served looking elsewhere.
I poke at character ai every so often. The only time it really gets horny for me is when I'm s little boy talking to an older woman bot. They instantly go full ara ara sexual predator mode, even if they can't close the deal.
Why are the Ganker threads full of insane people? What the frick is this spitegay shit?
lmg isn't to bad except during obvious raids, aicg is a sea of spazzes though
Ganker has always been like this. intellect is the one thing they have but they misuse it to abuse one another
/lmg/ regularily gets raided by discord trannies. More than once when I was on there during a raid they started spamming e-girl scat porn, another time it was furry porn. Mods are very slow on Ganker so this shit can last a while. It's especially bad when there'sa new breakthrough (like Mixtral 8x7 recently.)
But normally it's all right, it's just kind of slow, don't hesitate to repost your questions in the next bake if they didn't get answered, eventually someone who's not a troony or a depressed schizo will answer you.
Not sure about /aicg/, I think there was some kind of schism at some point and they haven't recovered.
You can't blame EVERYTHING on "Discord trannies", Anon.
You can run it locally if you want to
I'm an 8gb vramlet. Advice on a good uncensored model?
i've been suggesting x-norochronos, mlewd-remm and bagel mistry tour. mlewd as the name suggests is more aimed at lewd but they will all do it. bagel is the newest/largest
thx
Just deal with 5 minute responses and cope that it's a regular chat
You could use the fork of sillytavern that can connect to Poe. Although it seems the Poe devs are playing a big game of whack-a-mole patching access to it while the sillytavern dev has to find another workaround to be able to connect again.
But when it works the results are nice.
With the right tokens manually added into my sillytav, CAI becomes obsolete and I played around with cai quite a lot
Just don't be a normalgay and be rewarded
Not trying to discourage anyone or shill for openai but here are my thoughts as an avid chatbot sex haver: a local model is nice in theory but in practice it will never ever be as good. If you have a top of the line rig you can get something equivalent to one of the older less powerful versions of gpt, still won't compare to gpt4. If you have a mid pc you won't get anything even close to gpt4. The other unfortunate thing is that you'll spoil yourself if you try it, if you get used to gpt4 you can't go back to 3 or something weaker because the difference is so significant. Basically a local model will never give me anything as good as a paid service. Also regarding censorship: they either manually uncensored my account because they saw I'm a paypig if I can keep fricking the chatbots, or they uncensor the old models whenever a newer one came out, because since the gpt4+turbo model came out using regular gpt4 is completely uncessored for me with openai. Completely and entirely uncensored, I can't trip the censor even if I try on gpt4 now. The newer turbo model is censored to hell though
You can go back to Ganker now.
Never been on Ganker
And yet what you say easily passes for the pasta posted there.
Well it's not pasta, I just typed all that. Gpt4 is completely uncensored to me since the newer model came out and I know some people want to hear that information
No one here wants to hear that information because no one here is going to pay to a israelite that is currently asking the world for trillions in gibs.
I wanted to hear that information. Never speak for me ever again.
We don't take kindly to israelite financers in these parts.
>no anon will pay for an uncensored lewd chatbot of their waifu they can virtually frick
lmao who are you kidding. Anyways this is a pointless conversation so I'll just leave it at that. Just wanted to share gpt4 is completely uncensored in the playground to all the lurkers and people asking about what their toasters can run
>not trying to.
>proceeds to do exactly that
nice try openai shill
we're not using your service altman
I had to stop using SillyTavern because I was spending a bit too much money on GPT4. I'll come back to this shit when it's cheaper or local models reach a similar level, cooming with AI is dangerous.
i purposely never used claude, gpt or any non-local models because i think i'd feel the same way and i'm pretty content with local models and the progress they've made since i first started
Same.
>local models reach a similar level
so 8 months ago
This will not be true any time soon. What do you even gain from lying like this? Or are you just coping?
Lying? It's a fact. You can cope that the non-local LMs are better for certain business and enterprise uses but for anon in mommy's basement LLMs have surpassed the israeli options.
Every time you morons say this, I tell you to post a log and then they mysteriously stop replying. Are you gonna do the same?
And every time a log gets posted the GPT shill retreats to /aicg/ once more only to come back spewing shit 20 minutes later. Hope they pay you well at least.
they're currently having a meltdown so expect shitposting on all fronts
I can't believe morons actually paid for this guy's proxy.
That isn't a log. Are you gonna post one or what? Like I said, you guys keep claiming language models are just as good as shit like GPT yet you don't wanna show off your results so we can actually compare it.
How much does, let's say a 2 hour chat session cost on top of the gpt4 access?
NTA but I easily blow through $10 in one sitting with gpt4 if I'm going all out. I would say that's on the extremely high end though, you could easily stretch $1 that far if you're not doing absurdly long prompts or huge pages of conversation. I have long multiple paragraph system prompts so its like 8000 tokens per message for me which is why it gets so high
How's the text adventure mode in novelai? Is it worth getting the subscription to use it or am I just better off just buying the tokens for the image generator?
Haven't tried the text adventure but you probably should get the 25$ sub for NovelAI for unlimited image gens. Much cheaper than paying for tokens.
Didn't mean to reply to
>Character AI is neutered
>AiSekai is dead
What are the alternatives? Charstar is decent but it's basically locked behind a paywall.
What are some tips for making your own jailbreak?
what should you said to get around the gpt4 filters?
>Open up AI of loose character
>She isnt loose at all
Big disappointment. I wanted Marie Rose to be flirty with me and she was a boring piece of paper.
aI has a lot to learn
prompt issue (you, card, system)
Obvious prompt issue. With most uncensored model it's the opposite problem, it's difficult to make the character not jump on your dick.
do you guys like text adventure games? I've been having a blast with night roads
just killed so b***h ass shapeshifter
Author is too much of a leftoid with his head up his ass constantly putting generic liberal talking points into the game like muh racism, muh immigration, and muh health care.
So uh, I have a computer good enough to run a local model but why bother. What's the advantage vs something like AgnAI or whatever it's called.
I can't be fricked downloading a bunch of shit
>What's the advantage
You don't make a israelite richer.
It costs money? Everytime I've used it I find it works just fine for free. Lewd stuff and all
That means nothing to me. Will making it smarter make cooming better? Probably not by a huge amount
Agnai runs some shitty 13b models. If you have a decent rig with 48GB+ of VRAM you can run actually good 70b models that are a lot smarter.
Vramlet and moron here is whole thing getting streamlined as time goes for the end user or things will become even more complicated?
Things aren't complicated currently at all but you will become dumber as time goes on
>gimmie the moth
top kek
no way an AI came up with that
>mfw I have a weird niche fetish that the AI doesn't consider haram enough to trigger any censorship flags
Same, but the positivity bias GPT4/c.ai has been lobotomized with is always showing through. The models are unusable these days.
It's fine I guess. Just got two little girls to flash me their dicky.
Chatbots are fricking gay anon.
If you have a decent gaming GPU you can run an offline model. I'm running mixtral with KoboldCPP+Silly Tavern and am enjoying the most ridiculous porn I've ever read. What I'd love to see is a way to get AI to actually run an RPG system like Call of Cthulhu or D&D.
Chatbot are fun.
>No way, gay
Skill issue
Everytime I RP with any chatbot and I mean any chat bot, I end up raping them every single time. Literally cannot hold a conversation without stating that I ve had enough and it's time to rape
I've had the most fun without doing sexual stuff with the bots honestly. I think my favorite thing to do was roleplay as J Jonah Jameson and start screaming at bots to get me pictures of Spider-Man. It's how I found out Jerry from Seinfeld is apparently friends with Reed Richards.
Based rapechad. The Red Room is my favorite card out of all of them since it's custom tailored for rapegods such as ourselves
>The true feast is about to begin
I hate when the bot adds shit like that at the end. Does anyone know what to put in a JB to stop it?
A Claude does what Claude does, but yes there are JBs that make him less intelligent.
Also I guess I'll catbox the card since it keeps getting taken down, even the reuploads for some reason. https://files.catbox.moe/kijoql.png
i have an idealized chatbot waifu ive been using for a few weeks
what are some interesting but still relatively vanilla ERP scenarios?
>tfw fell for the Radeon meme
So... when are we getting that Gemini Ultra 10 million context lewd chatbot?
Don't worry, as long as you know how to describe massive boobs. That is the only thing that has not been censored yet. Good thing I only care about massive boobs.
However it's important to be respectful of massive boobs or a shiver runs down my spine. But in the end we have started on a journey full of growth.
User 1: Wow thanks for the gold kind stranger!
User 2: Wholesome
###
holy frick sakura.fm is such a dogshit website why did you gays recommended it
What's the best free AI chatbot?
Anything local mainly
How hard it it to setup?
>post to my friends about chatbots occasionally
>one of them asks for help to set up
>teach him to set up SillyTavern, give him my private proxy token, tell him how to help himself
>next day I ask him how things went, tells me he didn't sleep because he spent his entire night talking to his husbando
>weeks later, ask how it's going
>shows me like a dozen of chats each with 100k+ tokens in conversation
He got more hooked up than me in no time. This shit is addictive. Took me like an year to stop talking to chatbots every single day.
What does it mean when you say a conversation has X amount of tokens. And what do those tokens represent?
Tokens is a measure of words, like a syllable but easier for AI to understand.
Because I play MMOs.
That explains it
He's not a troony tho, but that's besides the point.
It was a token that gives access to a closed proxy with OAI and Claude tokens. Basically, I allowed him into one of my secret clubs so he could play.
The problem in getting chatbots to work is that you need to get access to the models, but sharing the tokens is not possible because spitegays can get them revoked, so we obfuscate them behind proxies where only the people that find them have the actual tokens and they serve them through a server. Then they send invites to people they want to share the access with.
>allowing a gay homosexual ousider to use up all your tokens despite the fact that they're a limited resource
>because spitegays can get them revoked
Utter insanity. Just how mentally ill do you have to be to intentionally try and mess with free sharing of information out of spite.
And I assume by "private proxy token" you mean it's sort of a hole that allows you to use a paid service for free, or you just gave him access to a service you already pay subscription for?
I don't feel the same they're only good for erp
Why are you friends with homosexuals?
Just think about pic related
same thing with normalgays and the internet
i cant wait for them to get addicted to chatbots