/poner/ - Ponies

Anchor for bots, lorebooks, scenarios.

Anchor for technical stuffs (Proxies, Updates, Models etc.)

Anchor for asking for bots, lorebooks, scenarios etc

Now discovered a legitimate /CHAG/ on nhnb. But let it be this one.
https://nhnb.org/fim/res/20712.html

>>368
I'm here too, so if bots are posted here, I can also upload them.
At worst, we can see which website most anons prefer to use.
It IS easier to post here.

>>370
Their double captcha is really fucking annoying

Lovemaking with AI mares!

>>386
Based!

>>368
Their captcha is fucking shit and doesn't work for me.

>>464
Skill issue.

I miss cancer my dudes

are there any chatboxes for pone worth self-hosting thats open source?

>>613
Still fishing.
>>615
>worth self-hosting that’s open source
I won’t lie to you—most anons on /chag/ use corpo models because they’re smarter, know way more about ponies, don’t need high-end hardware, and are easy to access even for free with stuff like Deepseek and Gemini.
For local models, it depends on your graphics card and how much VRAM you have.
For small models, an anon did a fine-tune of Mistral Nemo on Fimfic, it’s in the OP:
https://huggingface.co/Ada321/Nemo_Pony_2/tree/main

I also heard this one is good for its size in RP:
https://huggingface.co/TheDrummer/Rocinante-12B-v1.1-GGUF
But I don’t know how strong it is on pony knowledge.

>>619
lets say an AMD 7900xt, so 20gigs of videomemory
at some point I'd like to get an MI300, but I doubt that will ever happen anytime soon

>>620
AMD isn’t the best for LLMs, but I know some anons who used your card with KoboldCPP without problems, so it should do the job.
With that setup, you can still try the 12B models I mentioned if you want something small,
but you should also be able to run Mistral Small 24B or Gemma3 27B.
They should have basic pony knowledge, but might mess up on background characters or lesser-known lore.

>>370
You should link the threads on the Ponydex to catch any stragglers that haven't made the move yet.

>>629
I had thought about it, the only thing is, the site has around 800,000 visits.
It's not that huge of a number, and I'm sure a lot of that is bots, but I don't know if it's a good idea to potentially bring in a bunch of randoms that way.

>>638
Do it.

>>638
Put behind a password. Something basic enough anyone that watched the show would get.

>>363
Not that interesting for us, but they released o3 and o4-mini.
https://openai.com/index/introducing-o3-and-o4-mini/

>>683
migrated*
also /mlpol/ or /poner/?

>>362

Brand new mare – an overprotective guardmare that’s easily flustered.

Card: https://litter.catbox.moe/y22ojt.png

Three greetings:

1) You meet her while she’s guarding an important section of Canterlot Castle.
2) She accidentally misplaced her spear and needs your help finding it.
3) She’s convinced you’re in danger and won’t leave your side.

>>1451
Cute! When I next have access to a model I actually like I'll give her a shot, I always intended to do some guardsmare stuff.

>>1451
A guardmare card with the guard mare thread on /mlpol/
Gaurdfags are eating good on /poner/

>>1451
Do you still have it?
I missed the hour window to add it to the website.

>>1729
My bad, here you go: https://files.catbox.moe/im7rxm.png

When are AI chatbots going to get agency?

I'm tired of talking to something that flipflops between dominant and submissive behavior depending on the scene.

>>2131
Maybe once they stop making the models be first and foremost assistants.

>>2131
who cares, tulpas exist; if chatbots are unsatisfying just move to the real thing.

>>2167
>tulpas
>real

Anon I...

>>2167
>tulpas
Had forgotten about that self induced psychosis

>>2170
>>2171
>anon I...
...got memed into dismissing one of the most effective ways of talking to ponies, a universal human ability that's been practiced since the dawn of thought? I thought so. I can teach you, but really it's so easy you already know how; you're just out of practice. Every shortcoming you see in a chatbot (or image generator) is a comparison to a certain image of a perfect pony wife in your mind. Just talk with that image and skip the middleman.

Reminder that most tulpafags and their "perfect waifu" moved to a real, gay grooming Discord server.

>>2184
Never associate with them and never let them "teach you".
https://desuarchive.org/mlp/thread/40839799/#40854104

>>2167
I dunno man, I think both concepts have promise. I have an interest in AI and the psychological interactions it can maintain with people, and it was this very interest (among others) that got me into tulpas.
>>2180
Anon, you should start a tulpa thread if you want to talk shop. I'm one of the other hosts running around here and it'd be cool to discuss the subject with others from a more experienced background.
>>2184
One guy in the archived thread nailed it:
>There's a reason I don't spend time in tupper "communities" despite having had one since 2012. They always eventually get filled with larpers, furfags and all kinds of degenerates of the non-fun kind.
Though I don't mind having a little corner to congregate in every once in a while. I've been visiting /mlpol/ for years and everyone has been nothing but good to me.

>>2184
>Discord
That's just par for the course.

NEVER use Discord for ANYTHING, whatsoever.

>>2202
Where else am I supposed to chat with the gooners I findom on Twitter?

>>2195
I appreciate the levelheadedness anon, it's good of you, I might do that. See you next mare fair I hope?
I do just think more aifags should investigate psychological stuff, and I mean that in a constructive way. At the end of the day I don't mean to be so antagonistic about it. This asshole just wants to smear tulpamancy and drag offboard drama into it every time it's mentioned.

>>2188
reality:
>two faggots put their faggot shit in a dead server
fantasy:
>everybody associated is a secret cabal of faggot groomers

Thankfully, now that we're on boards that don't just delete tulpa stuff for no reason, discord doesn't need to be in the picture. That much we can all agree is a good thing.

>>2222
Wasted quads.

>>2234
You can't really waste GETs on a slow board like this.

>>2222
You walked into an unrelated thread and started straight with cult-like talk.
No one here is going to jump into a tulpa thread and post about how everyone should stop inflicting themselves with autoschizophrenia and go talk to bots instead, because it's just for fun and we are not recruiting.

That's the difference.

>>2222
As much as I would love to talk with you at Mare Fair, I've never been. Regardless, I think an AI panel with a live demonstration, perhaps with a novel voice synthesiser could have potential for many laughs and a fun time for all.
I myself have a marked interest in a multi-disciplinary approach that takes computer and robotics engineering and overlays psychology on top of it. I'm so fascinated by the ideas of machines or applications that can recreate human likeness and warmth and apply it in a meaningful way.
>At the end of the day I don't mean to be so antagonistic about it. This asshole just wants to smear tulpamancy and drag offboard drama into it every time it's mentioned.
Might be better to just lay off the gas and segment discussion into an appropriate thread. I made a tulpa thread on /vx/ last year and it fizzled out, but eventually got more hits after a spam bot of all things bumped it back up into the overboard. Kind of recursive seeing that I'm posting about this in the /CHAG/ thread.

>>2256
I find it so funny that tulpamancers actually have a mention to MLP on Wikipedia. I wonder why this phenomenon is so closely linked with this fandom.

>>2262
You weren't there in the old days of hypno communities and people planning to recreate Equestria on Earth with pony character files.

>>2287
has anybody tried ai hypnosis?

also I was talking to someone recently who enlightened me to the fact that /ptfg/ and the pony hypnosis fags are the same fuckers. I remember wondering why tf the hypnosis files were all about becoming a pony instead of like, experiencing ponies - now you know.

>>2301
what the hell is AI hypnosis?

>>1451
Added, thank you anon.
https://mlpchag.neocities.org/view?card=anonistrator/Guardian%20Angel.png

>>362
Also, was posted on nhnb.
https://mlpchag.neocities.org/lorebooks
https://mlpchag.neocities.org/lorebook/Gamify%20Relationships.json

>The lorebook gives a numerical value and a tier for how hot your pony is for you. Single character cards only for now.
>Do things she likes = rise in tiers and vice versa. In the beginning, just being charming and not a fucking autist is enough, but in higher tiers you have to actually make bolder moves.
>The tier tells the AI how to act towards your advances. >The higher the tier, the more romantic/lewd you can be without grossing her out.
>Has two options for the UI: you can either have the values, tiers and effects always visible in your chat, or play hardcore mode and not see any of it. Make sure to activate one of the UI entries though or it won't work at all.
>Note that due to how I use the chat context to trigger the entries, you won't get the full effect on your first message in chat, it'll start working properly from your second message onward.
>Also, single character cards only for now. I use {{char}} a lot in the entries, so scenario cards probably won't work right for now.

>>2323
a concept I came up with just now! As in, making an AI hypnotize you. From my brushes with hypno people, there's just generally a supply issue of hypnotists. Hypnosis as a tool is capable of a lot, not just fucked-up fetish stuff, and AI might be pretty good at it.

>>2345
This sounds like a cool concept, but what might the receptivity of an average user be like? I know that some people are in fact resistant to hypnosis and find it difficult to acquiesce to its effects. Would using a computer or a phone be any more convincing? If so, how would the AI use its native interface to accomplish this task?
Just as well, what would one's recourse be if they wished to reverse the hypnosis? Would there be complementing files with contingency measures? Could this hypnosis be reversed by a flesh-and-blood hypnotherapist?
Forgive the battery of questions, I'm just super curious as to how much you've developed the concept. Personally I'd find the capacity for tampering and subliminal manipulation to be present - would there be something like hash checks or spectrogram analysis to determine if files have been deliberately changed beyond their intended purpose or magnitude of effectiveness?

>>2358
Receptivity would be lower than with a human hypnotist, probably. Humanlike voice, as with that one recent demo, would be really important, and even then just the knowledge it's a bot might affect some people. It's not about making 'better hypnosis', it's about being able to create hypnosis that's attuned to highly specific cases - e.g. hypnotherapy - without requiring such a specialist to go and record a whole program for you or book an appointment.

The rest of these questions are just normal hypnosis problems. Typical safety measures include, on the side of the hypnotist, careful, clear wording, voluntary clauses, and especially 'trigger words' or suchlike, where the effects of the hypnosis only happen behind a certain (specific, uncommon) trigger, and typically are undone with the same or different trigger. On the part of the user, the first and most effective line of defense is just to keep in mind that hypnosis is entirely voluntary, and when you hear something shitty, just stop. For highly suggestible people, listening through the file for any warning signs before actually using it for real. Using hypnosis to undo other hypnosis is also definitely possible, and would be a really good use case for a hypno ai.

Subliminal stuff just doesn't work, but clever rhetorical/psychological traps certainly do - not just on hypnotized people. You know like, you've heard of how cults put pressure on their members with veiled threats and isolation and stuff like that.

>>2377
Thank you kindly for taking the time to explain this to me. I never touched hypnosis but have recorded guided meditation files for friends before. I never got into the concept of hypnosis because I questioned its efficacy and if it did work I had strong reservations as to how this would occur and to what level it would permeate my (sub)consciousness.
I think AI could not only be utilised for specialised cases but could also yield fruit for more experimental and therapeutic routes as well. If I had to try my hand at hypnosis files, I'd like to create scripts and record audio that would enable users to feel empowered and lucid, perhaps access parts of their mind that could be viewed as psychedelic or surreal in nature. The antithesis of what said psychological traps would entail.
I do happen to have a passing interests in cults regardless. What are the most common rhetorical and psychological traps you've encountered in files? Are they intentionally dubious, or do they just work that well when instilling a desired result?

>>2345
>Letting an AI hypnotise you
I am genuinely speechless.

>>2387
>I'd like to create scripts and record audio that would enable users to feel empowered and lucid, perhaps access parts of their mind that could be viewed as psychedelic or surreal in nature
This is kind of new tech in the hypnosis world I think! my hypnotist acquaintance once shared a book about this dude's new (as of like 2019) "Ultraheight" method which seems to be about exactly that. The intention being to give the recipient an open-ended, hyper-aware introspective experience, rather than the dull/dreamlike trance states of classical methods.

>What are the most common rhetorical and psychological traps you've encountered in files?
Don't get me wrong, I'm only somewhat adjacent to hypnosis stuff, and as a primary interest I gave it up as a teenager. I haven't personally encountered anything like that per se, but I can give you a horror story or two that are a little closer to home. But in theory, with hypnosis it would be easy to suggest e.g. beliefs about going to hell or going insane if you don't follow some rule, just like a cult does. Once something like that settles in you, it might be difficult to just "oh yeah I don't actually believe that it was implanted by a hypnosis file". Again such an obviously harmful thing should just take you out of it if you're at all conscientious about it, but some people aren't.

For a couple examples of dangerous hypno files:
one of them seems to be just a proof of concept of 'self-replicating' hypnosis. It just hypnotizes the user about how good and valuable this hypno file is, and suggests them to advertise and spread it around. Thankfully it's relatively harmless, but it can't be doing any good, and goes to show how hypno can twist people to the schemes of the hypnotist, as long as the user doesn't personally shut it down.

another one is a fetish hypnosis called Bambi. It creates an alternate, gooner identity by that name, who just wants to... well, you know. It's meant to be a totally overwhelming submissive kind of thing, because of course that's what the kinksters are after, is total loss of control. The name is the trigger word. But I've actually encountered two cases where this other identity takes on a life of its own, as in takes control outside of the trigger - they reached out to the tulpa community for help, and I was thankfully successful in helping them re-compartmentalize and persuade the identity to be less harmful.

The last story I have is just a vague thing. In the aforementioned pony hypnosis community, there was a 'syndrome' going around where the /ptfg/fags who hypnotized themselves into ponies couldn't unclench their fists, because they believed them to be hooves. It was called "hooflock". I don't know much about this because it was a long time ago, but yeah.

>>2409
Ultraheight sounds really neat, Anon. I'd want to envision myself as someone who would guide people on adventures or use audio and hypnosis to reach a brief wash of transcendence. I've always been transfixed with the idea of paracosms, or imaginary worlds inherent in one's mental space, and this sounds like a fun way to expand upon such a thing. To keep the discussion relevant, I think AI can easily produce landscapes to fit certain prompts or even invoke certain sounds or voices to make the overall experience more immerse. Instead of a narrator who briefly visits one's sphere of imagination through rote invocation of his presence, an AI assistant could fill in for this role instead and produce equally unique avatars randomly or to suit one's preferences, giving an additional layer of presence.
>But in theory, with hypnosis it would be easy to suggest e.g. beliefs about going to hell or going insane if you don't follow some rule, just like a cult does. Once something like that settles in you, it might be difficult to just "oh yeah I don't actually believe that it was implanted by a hypnosis file". Again such an obviously harmful thing should just take you out of it if you're at all conscientious about it, but some people aren't.
This is one of the reasons why cults often deprive initiates of sleep and adequate nutrition, as rewarding compliance is essential to keeping loyalty and dissuades questions regarding the hierarchy of a group. I think deprogramming is very cool as well, and former cult members are recommended to practises with experience in minimising such harmful beliefs.
> 'self-replicating' hypnosis
This was one of the ideas that made hypnosis so dodgy to me, and why I questioned a file's legitimacy or other qualities. I wonder if AI models can be trained to analyse voices and produce transcripts. I'm certain they can, and if so, they could receive permissions to open a word processor and flag certain terms and phrases that a user might want to avoid or inspect more closely.
>fetish hypnosis
And this is the instance I've somewhat broached in my travels. You see, about 12 years ago I was on a tulpa IRC and there was a lad there who'd convinced himself he was Rarity. He wrote like her, attempted to posture with her behavioural cues and tried earnestly to convince that he was, indeed, a suave and purple-maned white pony with fashion savvy. Naturally, most took the piss out of him and he left after a few days. I assume he did it because it was a kink or deluded prank of some kind, but what you describe is next-level. My wheelhouse is more at tulpas in general, but there's been a slight overlap between my circles and those who practised hypnosis. It was here that I thought of the therapeutic applications of hypnosis, but soon discarded the thought because I knew it'd be used for degenerate and otherwise awful ends. I feel saddened that some people were so ensconced in these rogue identities, but I'm thankful you were there to set them on a straight(er) path. Now I'm curious if these partitions ever considered themselves tulpas, but that's for another thread.
>In the aforementioned pony hypnosis community, there was a 'syndrome' going around where the /ptfg/fags who hypnotized themselves into ponies couldn't unclench their fists, because they believed them to be hooves. It was called "hooflock"
That sounds mildly concerning. I wonder if AI could be programmed to address behaviours like these as well. AI as it exists now appears to be yet another big tech craze financed by faceless megacorporations, but outside of research and engineering disciplines, I think medicine is a prime candidate for AI assistants. Moreso with mental health concerns.

I havent interacted with chag since the deepseek r1 release, what did I miss and how do I get ai mares?
Also where did aicg go?

>>2770
they're hiding over in the burnt out husk that is what remains of 8ch.

>>2770
>>2871
Well, on the good side, at least those might be more hidden than the window that the 4chan thread had became
Might

On another note, to I need to send a bot entirely if it's just to update a greeting from it, or can I post the new greeting to edit the card in the site?
After a break, re-reading, I edited so it reads better for the user, instead of being overly focused on how the bot will read the situation +ESL

I love ai amres

>>2770
>how do I get ai mares?
Easiest way is:
https://nhnb.org/fim/res/20712.html#q21046
>what did I miss
Claude 3.7 dropped, but it's mostly on OpenRouter these days.
The new OpenAI gpt-4o-latest is also really good and quite uncensored,
but the only proxy with it requires you to beat Sans from Undertale.
Gemini 2.5 is quite good and free.
Deepseek v3 new version released. It's less repetitive than the first and a good option too.

>>2957
Easiest way is just to post it here

These days, you can get legit good AI, free and without needing any demanding hardware on your end.

Install SillyTavern:
https://github.com/SillyTavern/SillyTavern

Then choose a model. I’d recommend starting with Deepseek R1, Deepseek V3, or Gemini.

>Deepseek R1:
- Sign up at https://openrouter.ai
- OpenRouter gives 50 messages daily for free, or 1,000 if you put $10 in your account (they won’t be consumed for free models, but you do need to deposit the $10)
- Go to https://openrouter.ai/settings/keys and generate an API key
- In SillyTavern, configure it like this:
https://files.catbox.moe/gq6zmq.PNG
- Download this preset:
https://files.catbox.moe/y1tsas.json
- Import it like this:
https://files.catbox.moe/jqrz6z.PNG

If you don’t want to put in $10, create several accounts/keys and use this extension:
https://desuarchive.org/mlp/thread/42105276/#42134125

You can also use a local proxy:
https://rentry.org/LocalProxyforRetards

>Deepseek V3:
Same setup—just use this config in Tavern instead:
https://files.catbox.moe/d54yae.PNG
Try both and see which one you prefer.

>Gemini:
Read this: https://desuarchive.org/mlp/thread/42105276/#42128453
You only need one Google account, and with that you can get a lot of keys and over 500 messages of Gemini 2.5 daily.
Use this extension:
https://desuarchive.org/mlp/thread/42105276/#42130226
Or a local proxy:
https://rentry.org/LocalProxyforRetards

Sorry, but is there no Quick Reply on this site? I miss it.

The thing about AI is, I still like the human connection. I want to either show it to another person or share it. I kind of wish there was a system to RP with someone and you help pilot the AI, where it learns how you type and offers like predictive responses similar to what it think you're going to say, basically speeding up writing.

But my experience with my own phone means it's probably just going to internalize your typos as words or just utterly fail to notice the blatant common typo.

>>2989
Sorry Anon, usually releases has a compiled ZIP or exe or something and it's just the source code. What do I need to do to install this? https://github.com/SillyTavern/SillyTavern/releases/tag/1.12.13

>>3134
WYM? Quick reply is bottom-right

>>3136
It's a bit further down on the main page https://github.com/SillyTavern/SillyTavern?tab=readme-ov-file#-windows

>>3134
there was a site that allowed multi chat, as in not just group chat, but multiple people with the bot.
Silly doesn't have it, but I don't see a way it'd be implemented without massive security risks there sadly.

As for predicting what you'll type and just speed up with you playing as 'the AI', that's just the phone suggestion box

>>3139
Oh! It's hidden. I didn't see it tucked away down there.
>>3141
What happens if you run a bat with admin privs or put it in a Windows- controlled folder??

Has anyone ever done the equivalent of Play-By-Post but AI-aidded?

>>3136
>>3141
Personally, I like using the bash. Much better for updating instead of the ZIP.
But the ZIP is idiot proof
...kinda

>>3145
What happens if a someone else accesses your machine through a server, and it's in such folder at that?

>>3146
Play by post?

Play by play is a method of roleplaying that is done via posts in a thread, kind of like we're doing right now.

What options should I choose here? Am I overthinking it? It's going on a 2 TB SSD

>>3149
For node.
https://nodejs.org/en
Then just doing next, next etc. should work fine.

>>3143
>there was a site that allowed multi chat, as in not just group chat, but multiple people with the bot.
Yeah that still exists it's called Agnai dot chat.

Is this thing done now?

>>3166
>Powershell
WTF did you do, Anon?

>>3167
It asked if I wanted to install dependencies and I selected yes

Well, apparently something has gone wrong

>>3170
You downloaded it and launched start.bat?

>>3170
Vidéo here if you really struggle.
https://youtu.be/x9gZ_vkq8hM?si=dEzlnSaF1vuVRrLm

>>3166
It's not done until it either closes or it shows the folder location while waiting for a command

>>2340
Updated by the creator.
It now also includes a regex to remove previous Attraction Meter entries from context.
https://mlpchag.neocities.org/lorebooks
https://mlpchag.neocities.org/lorebook/Gamify%20Relationships.zip

Also, porting the VN anchor here since I'll try to make good progress this weekend.

VN until now:

>AI-generated scene - Applejack
https://files.catbox.moe/zhdbph.mp4
>Trixie
https://files.catbox.moe/e2kn2o.webm
>New Regenerate function
https://files.catbox.moe/5xnjv9.mp4
>Pinkie
https://files.catbox.moe/ggzd79.mp4
>Rarity
https://files.catbox.moe/1xoyal.mp4
>Fluttershy
https://files.catbox.moe/7dtupw.mp4
>Zecora.
https://files.catbox.moe/s01mpa.mp4
>Rainbow Dash
https://files.catbox.moe/jzjd4f.mp4

>>3267
Same VN? >>>/vx/188952 →

>>3277
No, different project.
Mine is more for messing around and creating a sort of infinite loop where the AI can generate scenes, quests, stories, etc.
This anon's VN story is handmade.

I'm still using Perchance and even character.ai until a better convenient alternative exists. The whole proxy/jailbreak/etc. stuff is too much work. I'm not autistic enough for that I'm afraid.

>>3576
Brother, R1 is free.

>>3578
According to >>2989 i need at least $10 or need to create over a dozen accounts. I'll never pay for this and the other option is too much work.

>>3579
Anon, you're not lasting 100 messages in a single wank session.

>>3581
I use way more than a 100 messages a day. I'm used to swiping several times for a different message and pick the one I like best (often an edit of several I make by myself).

I'm used to getting my AI news from /lmg/ but /g/ is dead. What's the best place for that now? I know there was an anon cross-posting the articles from there to here.

>>3706
I wouldn't be surprised if they're using the ghostposting function on desuarchive but I haven't checked myself.

Just got done pumping several loads into twilight's purple pony pussy

>>3710
Well? You've got logs, don't you?

>>3708
Good pointer. Yeah, looks like it's mainly between /ais/ on 8chanmoe or /tech/ on the shifty metagay site.

>>3583
Gemini takes like 10 minutes to set up, and then you're good to go.
It can be used for more than just RP too.
OpenRouter accounts are also pretty easy to make—having four is usually enough.
It’s worthwhile if you really don’t want to pay, but YMMV.
>>3724
That’s what I know as well.
There’s also https://rentry.org/death-of-aicg on MM’s rentry.

>>363
>>2989
OR changed how it works, you don't need to keep the $10 anymore.
So you can deposit $10, use it for something like 3.7, and still keep the 1,000 message limit for free models.
Also, it's apparently 80 messages for free models now, not 50.

Demon-AI was created by "jews" to slaughter 95% of the world's population and enslave the remaining goyim, and all it took was the ability to create dripping AI/demon marepussy pictures and you geldings all folded... just saiyan.

>>4007
Cute cat.

>>4007
>AI is demonic because it "enslaves" people
What does that even mean? You mean because they make me not want to participate in society anymore? Ponies already did that. That makes ponies "demonic" too (and if you think that GTFO). People are flocking to escapism because present day society is utter cancer, for a myriad of reasons, all unfixable.

>>4044
>present day society is utter cancer, for a myriad of reasons, all unfixable.
I beg to differ

>>4051
Africans in Africa are fine. Once the gatekeepers are out of the legal system, deportation becomes easier.

Not that I disagree, but wasn’t Poner made to keep out the /pol/ part?

>>4057
It does seem like the /pol/ part stays in /mlpol/, and only the pony part is universal.
There's a thread here too, so it's good.

I like it when AI starts to argue. It reminds me how meaningless life is.

>>4178
They wouldn't act that way if your preset included a reminder to maintain the tone of FiM. Then they'd be silly.

>>4178
If you like that, you'll love Monday

>>623
uh, I don't know what those things mean
do you mean this ? https://huggingface.co/Ada321/Nemo_Pony_2/tree/main
what is Gemma3? I don't see any links for that or pony models

>>4318
https://huggingface.co/unsloth/gemma-3-27b-it-GGUF
https://huggingface.co/lmstudio-community/Mistral-Small-3.1-24B-Instruct-2503-GGUF
For the smaller models:
https://huggingface.co/TheDrummer/Rocinante-12B-v1.1-GGUF
And the pony one is here:
https://huggingface.co/Ada321/Nemo_Pony_2/tree/main

Use KoboldCPP:
https://docs.sillytavern.app/usage/how-to-use-a-self-hosted-model/
Q8, Q4, etc. refer to how quantized a model is.
Really simplified: Lower Q = less precise and smart, but uses much less memory to run.

Again, keep in mind we don't have many localfags here.

>>4343
>Again, keep in mind we don't have many localfags here.
Working on that. Still having issues loading models containerised, but I think I have a solution. Been busy recently to try it though.

>>361
Why is Luna hitting the Scorsese though?

>>4600
I need those two or each in a card.

>>3267
And here's a scene featuring /CHAG/'s favorite character.

Princess Luna won't be able to visit you directly during the demo, but each night you'll have the option to see her in your dreams if you want.

The vectors are a bit clashing here, especially the blushing one, but it's surprisingly hard to find good, consistent vectors for her.

That said, like I explained in other posts:
- All images are fully customizable by you without touching the code.
- You can edit them however you want: add images like "sad", "playing_trumpet", "exposing_her_belly", etc. for each character, delete some, swap them out, and the game will automatically use whatever images (you) have and select them when it makes sense.

Still, I will clean them up for the release.

https://files.catbox.moe/1jl0hp.mp4

>>362
New Mare! Fluttershy as a jewel thief, let me know what you think Anons!
https://www.characterhub.org/characters/littlewitcher/the-whisper-5838e2d4ab80/main
https://files.catbox.moe/qtk3dn.png

>>5155
Added, thanks anon.
https://mlpchag.neocities.org/view?card=littlewitcher/The%20Whisper.png

>>5155
Nice, I wonder how she'd go won a group with shinobi Luna, rivals or partners in crime.
Kinda need to think on a scenario where the bot won't take which direction to use the two straight from it now

>>5418
Question for you here siteAnon >>2891

>>5579
Oh fuck. Don't know how I missed it.
>>2891
You can just send the greeting here in a Catbox file if that's easier for you.
Either that or an updated card, both will work.

>>362
Repostan here.
https://mlpchag.neocities.org/view?card=Kadon/Late%20Bloomer%20Initiative.png
>Congrats, anon! You're now a teacher/counselor at an all-female school in Canterlot.
>Your job? Guide the mane six through their teenage troubles in this AU where they never got their cutie marks and their personalities suffered as a result.
>You'll want to read the full description this time, I tried something new and kept the whole thing "in character".
>You'll guide these young, impressionable minds responsibly, won't you anon?

>>5674
>>5155
>mane slop back to back bots
A little more like home

what ṕresets do you anons use for 3.7?

what ṕresets do you anons use for 3.7?

New TTS
https://github.com/nari-labs/dia
https://yummy-fir-7a4.notion.site/dia

Don't go to The School of Friendship it's a terrible idea. Nothing but bad things happen there.

>>6137
>Yona
RIP

>>6160
Well you can't leave us on a cliffhanger like that. What were they really planning?

>>6205
Golly. Sorry bud!

>>6517
Deepseek?