/poner/ - Ponies

>>6670
It's https://perchance.org/
I’ve heard it’s not that bad for an online chat service like that,
but yeah, it's hard to help since not many anons here use it.

His preset doesn’t sound that bad for short outputs, at least at first glance.

>>6553
>>362
Both updated, thanks anon.
https://mlpchag.neocities.org/view?card=Anon%20and%20Anon%20accessories/Cloudchaser.png
https://mlpchag.neocities.org/view?card=Anon%20and%20Anon%20accessories/Cherry%20Berry.png

>>362
Fallout: Equestria Tactical Game

Random quest, random weapons, random team of clumsy-hoofs losers, and many dangers on the way.

Stories from my previous games:
1. During the first game with team of two mares we stormed The Radio Tower to fix repeater and accidentally burn The Radio of Metal Raiders. The next day we make a deal with doctor of nearest town. Escort our caravan to home turned out to be surprisingly difficult, the Radio Raiders were really mad, but fillies from Stable with flamethrower truck saved us.
2. Together with scout-filly we found a gang of slavers with their own fucking locomotive, and when they have fun on railroads captured their settlement under the railroad bridge. Filly with sniper-rifle on the beams of bridge and some distraction works just perfectly against half-dozen of tipsy guards. With liberated zebras we finally ambush the train, when they returned. We just use explosives from nearest mine and ballast as shrapnel.
3. With one pistol and one crossbow we fought against rednecks for desert lake. With their rusty shotguns and some crazy engineering skills we made our way throughout undergrounds of our malfunction Stable and found prewar labs of Forcefield Project. This story almost ended on the bridge above reactor, when sentient tank-bot catch us, but fortunately we hacked the door of armory and found set of power armor and two railguns just minutes before.
4. Just recently with a filly named Gears we used surveillance cameras, self-made anti-personnel mines, electric fence and 'Debris Scraper 3000' to secure existence of our species and our water-power plant against herd of rad-deer-critters.

Finally, I can post from TOR in my beloved /chag/ and share my cards. I make games for small local models, like Llama and Mistral. So its very interesting, how this card works for better models. And if someone edit this card I will be grateful (English is not my language).

>>6801
JSON version:
https://paste.debian.net/plain/1371200
/poner/ does not like cards.

>>6801
Nice, that's amarezing. I'll certainly be checking this out. I kept using Littlepip and sometimes Blackjack for these, having a proper sim to add mares at times is like a wish come true. Thanks Anon.

>>6807
Also you can use catbox to send pngs if they're stripped like they were back in 4chan, but json is fine too.

>>6807
Probably strips the file exif just like on 4c

>>6979
It also fucks the image up in the process, or at least it did for one of my pictures, I haven't tried any others. I tried to upload it to the booru, which rejected the image with a "width/height must be greater than 0" message. I had to first run it through jpegtran to be able to upload it.

>>6801
Added, thanks anon.
https://mlpchag.neocities.org/view?card=Anonymous/Fallout%20Equestria.png

>>6801
>>7186
Local models are good with F:E Lore?

>>7189
Depends if they were trained with ponies. Some of the pony models might, but doubt general would have much in local ones

>>7186
Thanks you too, anon. My first cards were from mlpchag, so it's nice to see my contribution there.

>>7189
No, overall they are bad with detailed lore, but surprisingly good with crossovers and battle scenes.

I usually use 8B Llama, 12B Mistral, 12B Magnum. All small models are quantized too much, even if they trained with FoE fanfiction, without lorebook they don't understand major events like permanently cloudy skies. But they perfectly understand concept of Fallout and Equestria crossover (I have never met humans in my games).

Instead of token-heavy descriptions and lorebooks I used one quite effective trick for prompting. TF-IDF analysis of FoE compared to 100 random books. This creates a short list of characteristic n-grams (rare phrases of 2-3 words), which works like keys to desired atmosphere, themes and setting. Small models very sensitive to that n-grams regardless of context.

Fallout universe n-grams (basic set):
"nomadic gangs"
"ghouls in ruins"
"Vault Door"

Fallout New Vegas / Western n-grams (expanded set):
"oases of civilization"
"Wild West"
"Rocky Mountains"

Fallout Equestria n-grams (crossover set):
"magical radiation"
"Equestria and Zebrica"
"The Stable"

Military n-grams (make enemies clever):
"Great War"
"small arms and light weapons"
"radio communication"

This phrases must be rare generally, but typical for thematic texts. And then the magic happens, LLM addresses thematic knowledge much more often and in the desired combinations. TF-IDF method allows quantitative estimates for prompting before in-game tests.

>>7186
Nice

/mlp/ is back

>>7796
Captcha doesn't work

>>7803
Works fine on my end. Word of warning on those, I have to wait a flat 2 minutes every message so we may be sticking around here for a bit

>>362
From the other bunker thread:

>Been on a card creating binge. Have another. This one was inspired by littlewitchers jewel thief Fluttershy.
https://www.characterhub.org/characters/Kadon/the-fillydelphia-four-87f4ba268a9c
>Introducing the Fillydelphia Four, Equestrias best heisting crew. Includes three greetings:
>(you) get taken as a hostage after a job gone wrong
>plan a job with your newly hired crew
>go on your first job as a potential member as the mares test your mettle

https://mlpchag.neocities.org/view?card=Kadon/The%20Fillydelphia%20Four.png

>>7828
4chan is sort of back. Captcha can still be a bit fucky, but it's possible to post.
I'll keep lurking both bunker threads to watch activity here, but I’ve got no doubt most of it will return to the main site.
Still, these are good places, and having somewhere a bit less public can be nice for this hobby.

>>7829
>but I’ve got no doubt most of it will return to the main site.
I don't mind staying here.

It looks like they are still testing things.

>>7833
Makes sense.

For the logs posted on both bunkers, should I just do one recap post with catbox links or just post all of them on different posts? Second one is a bit spammy, I guess, but I like logs and they are important IMO, so I don't care that much.

https://poll.horse/KApDCXNp

>>6801
>>7359
Thanks again for the card, anon.
I don't know if the thread will survive here now, but it's a shame we can't have you in the 'main' one.

>>7937
>"""main""" one
Reminder that with that you basically mean /mlp/, which is also back.
Honestly I kinda don't want to go back to there, but I'll keep lurking around

>>6801
Instead of a horse, I took the json and decided to use this image instead, felt more fitting for a ambient card
https://ponerpics.org/2100826

Making the town hall center fit in the middle of the rules of third bottom right or so

>>7829
>>7937
I'm stay. Just don't want to use ssh-tunnel for 4chan again. Too much efforts, I even don't have expendable vps new. Furthermore /poner/ works just perfectly without captcha and js.

But I still have few cards like Fallout Equestria to translate. And some information about preset-making for our chag-bros.

>>7960
Why not? I like this image too. I just took random mare with duster for card from my /foe/ folder.

>>363
I noticed that /chag/ and /aicg/ anons still don't know about brand new zigger technology, The NoAss Preset:
https://rentry.org/NoAss-GigaSchizoKostyl-Redux
https://rentry.org/noass_ext
English anons sees it just as separate avatars for multi-character cards, but this is much more complex idea.

Our /aicg/ discovered that usual chat-format disturb the creativity and initiative of simulation. Too many limits. Model forbidden to reply as user and his mares in group chat. Model sees post as separated objects and often interpretating user's words as instructions. Overall model sees narration rather like a chat than like a fiction. But best parts of datasets for LLM mostly consist of books!

NoAss, mean 'no assistant'. Model is not assistant. Instead of usual chat this preset allows model to generate endless story. Instead of chat-like posts we use stop sequences for user and other characters. Screenshot is typical example, narration from simulation, then chain of injections from user and his catgirl. But simulation sees user and his catgirl as it's own characters inside story and can completely ignore user's accidental instructions. NoAss preset trick simulation to see user as character inside fiction.

Some schizo-terminology:
- FullAss -- united unificated narrative. Simulation narrate endless story without any control and interruptions by user.
- NoAss -- united not-unificated narrative. Simulation narrate endless story with injections from user and other models.
- HalfAss -- multi-turn unificated narrative. Simulation narrate story, but time to time stops and suggest decisions from user. Like COYA.
- Ass -- multi-turn not-unificated narrative. Simulation narrate story in chat-format and react to posts from user and other models in group chat.

We have already used almost all possible tricks to make Ass better. But ass is just ass. If we look at the LLM logic and output more comprehensively, this is still a lot of possibilities to improvements, if we remove one of deeply hidden unnecessary limits. We can create smooth group chat without separation of narration. We do not need LLM assistance for multi-turn narration. From recent months 2ch /aicg/ slowly drifted from Ass and HalfAss to NoAss format. Finally, anon created Redux Preset for Silly Tavern users. With avatars and much more functions inside.

Try NoAss, bros, this is completely different experience even for well known cards.

>>8011
Since you're not talking to a single mare but meant to be a scenario.
Just an autism of mine

>>8014
>NoAss
Placebo. I tried to use it in non-pony RP, but it persistently put all sorts of mutant-type bullshit or mysticism where it shouldn't be.

From the Russian presets I took remix poluzhop.

>>8011
Alright poner, we better see some zala logs if you're staying, You heard the mods didn't you? Right of passage,

>>8063
>You have a right of passage to do blur blur
Do you even know how many guards Zala has fugged? Do you know her lore?

Sorry anon, it's the rules.

>>8064
Skill issue, AU it up and take a Zala who's never been with anyone and a Zala who's only been with stallions then convert her to HMD

>>8062
>Placebo. I tried to use it in non-pony RP, but it persistently put all sorts of mutant-type bullshit or mysticism where it shouldn't be.
NoAss is very demanding to Greeting and formatting. Each card need to rewrite, but script "Rewrite greeting" from NoassExt-utils.json almost useless. In fact, we haven't found consensus about best style of formatting yet, but undoubtedly correct greeting is much more important than with usual ass-prompting.

Try theatrical way of formatting:
1. Act/chapter/location -- first thing in Greeting.
2. Description of location, characters -- second thing.
3. Short example of user's speech and actions -- third thing.
4. Example of location change / event / timeskip -- between.
5. Short example of interactions with another char -- at the end.

>poluzhop
Lol, this literally means HalfAss. I still prefer HalfAss presets too, because of small models, but NoAss looks much more promising, even if it just first steps.

General opinion:
Best for HalfAss -- Gemini
Best for NoAss -- Sonnet

>>8063
You could just admit that you want new card with Zala, hehe. Okay, challenge accepted.


[NAUGHTINESS] - mental state, represents naughty temper. Initially 0%. Increased by daring acts, especially when Zala is in control. Increases her bravery to impressive level. After 25% she is ready for…


You know, small models like growing stats. Quite slowly, or not so slowly growing stats.

PS: Speaking seriously, I'm glad, that /mlpol/ still cozy and alive. I missed the birthday event itself, but two years ago found archives and read for hours with wide smile, even without most of images. Mega-impact. Literally.
>Two boards that 4chan did want to end
>Instead became much dear friends

>>8080
>You could just admit that you want new card with Zala, hehe. Okay, challenge accepted.
I was just thinking the same thing, but for >>8064

She hasn't gotten updated in 2 years and deserves a little better. /mlp/ never went anywhere with co-labbing cards but maybe /poner/ can do better. Fork her, new intros, a rework, not the slop nstdude left us with. I know that anons on /mlp/ turned up their nose at ai assissted intros or ai genned intros. But LLMs are a little better now, what's the opinion on it? I don't think it would be any worse off than the tiny into she already has.

>>8146
The card is old, yeah. It was still Card V1 back then too, so only one greeting was possible. And ntsdude was only doing NSFW.
>What's the opinion on it?
Anons tend to turn their noses up at fully generated AI greetings that are easily identifiable as such. That means:
- Lots of "isms"
- Sentence structures, rhythm, and patterns that are clearly AI-generated
Because they can often influence future outputs and contaminate the interaction right from the start.

Without effort or knowledge of what AI can produce, it's still easy to end up with something subpar like that, even with today's models.
If you do something:
- That is AI-assisted but with human writing as the draft, or
- AI-assisted but not easily identifiable as AI

I don't believe anyone will complain.

>>8167
What would be some isms, structures, patterns that should be avoided?

>>8186
Depends on the model.
You don't see them as much nowadays, but things like:
>testament of
>mix/mixture of X and Y
>too much focus on eyes/gaze, etc.
>mischievous smile or mischievous something
>ministrations
>maybe just maybe
>Somewhere a X (Deepseek)
>your greeting starts with the sun going up, down, illuminating, etc.

For sentence structure, it's mostly about avoiding all your sentences being like:
>{{char}} X "Dialogue." She Y, her voice/eyes Z "Dialogue."
>{{char}}'s eyes/voice/etc. X "Dialogue." She Y, etc.
They are basic sentence structures, so it's normal to have some of them, but certain AIs can spam them a bit too much.

There is also '—' spam on some models etc.

>>8167
I strongly prefer to made greeting by myself, despite bad english. Because of bad english.

Card is finished except tests and final editing. My personal filly for cum and amusement, kinky storytelling and silly-tuned political games. Alas, my Zala still too coy for her expected behavior. I mean, to rape poor guards in her particular innocently seductive style. Although I have not yet reached even 50% Naughtiness, at current level she is still too far away even from Cozy level of yangire-psychotype.

Any ideas? Any additions to her lore? Full card, if someone really interested:
https://ponepaste.org/raw/10923
Work in progress, final version for mlpchag after some tests (maybe 2-4 days). In other words I'm first.

>>8186
>>8191
Actually, this patterns can be found by TF-IDF method, which I mention before.

Every LLM, just like every human being, has very individual writing style (preferences in vocabulary). This style can be detected just with 10k word sample with one of my tools (less than 200 man-hours). For example, this is similarity detection, groups of texts from finfiction-like site. Even when I remove first 50 rarest n-grams from every list (usually names, locations), this works still perfectly and detect author in database of thousands of texts. I think, modern tools can detect authorship just by 1000 word sample or even less.

If you provide at least 10k words output from your model (better FullAss type of narrative from different cards), I can create list of that patterns. This is simple, just bare TF-IDF to compare LLM output to 100 random books. Besides, because it is simple, I'm sure someone made that type of lists already for every model. Try some search first. 4chan archive, github and scientific publications. Keywords "TF-IDF", "2-grams", "3-grams".

>>8191
>There is also '—' spam on some models etc.
Looks like influence of flibusta library dataset. This is our national way to separate complex dialogues and narration. Much better than "spruce forests" I must say.

>>8602
What the fuck am i looking at. I can't understand vodka runes

>>8690
>vodka runes
This is book titles. Model used central book to find all similar books in library, as result we see all stories of that author and others, who write for that setting. Second example without first 50 terms (character names, locations, setting-related terminology). Gedzerath and PhoenixDev whrite for one setting, but their writhing style completely different. This is different persons.

How it works:
1. Dataset consists of n-grams (words, 2-words, 3-words phrases). Separated lists for each book.
2. Frequently counter make dictionaries from lists. key:value, where value is number of n-grams in list.
3. Percentage: (number of n-gram / sum n-grams in list) = united format for different books.
4. TF-IDF: logarithm (number of n-grams / number of lists with that n-gram) = rarity of n-gram
5. Vector space. Each dictionary as node, rare n-grams as connections, similar values as connection weights.
6. Simplification. Each node to each other node. Only strongest connections will survive after each step.
7. Database: nodes as dictionaries. Keys -- links to other nodes. Values -- pair of n-grams and weights.
8. Summarize: sum of all survived connections as weight between nodes.
9. Format: graph of nodes, connections and weights between them.
10. Output: graphviz.

From first steps this works almost like preparations for new dataset to train LLM, but on "Vector space" stage we used various methods of simplification to study anatomy of the LLM during training. On "Summarize" stage we can see skeleton of LLM, strongest connections and some remnants of internal structure. I used it to make similarity detector for authors in dataset.

This is my own method, nothing professional. I just writefag with some knowledge about coding and old-fashioned syntactic analysis. I even don't know most of modern terminology, just like to do my own research. But results still impressive.

>>8014
So, turns out I've been doing NoAss for a while. I've always had the irrational distaste for the assistant because I want to make minor edits all over the place, and the perspective shifts trip me up, and make things just not flow.
The techniques in the rentry are more extensive than what I was doing, but they're consistent.

However, I'm pretty certain it won't unlock any power with existing models.
This is evidenced by the effort they have to go through to undo the assistant finetuning, and the loss of zero-shot prompting. Basically, the perfectly fine literaturary ability of the LLM has already been partially lobotomized to make the unnatural user-assistant writing style the default.
A particularly egregious finetune might use this technique to pull the LLM into an entirely distinct, 2nd-person-perspective vectorspace. Most models are not this bad though, especially models that advertise storytelling capabilities, and won't see much change in capability.

It'd be another matter if you were using base models, or managed to find a model that was not tuned as an instruct or chat. These have always been notoriously bad at instruct and chat, exactly because they're naturally NoAss.

>>8080
Some people like gamification cards too. Like win/fail states What would be something like that for the little zeeb?

Deepchink has 2 new models, anyone try them?
Chimera is supposed to be good, but I haven' t had any luck getting it to generate anything.

>>8951
Yeah, interfaces. Actually, I hate this part for cards with mares. But this helped a lot for small models, you can reroll bad post earlier, or edit just first part to control following narration. One or two stats + some simple planner helps to control the narration and prevent some loops. Usually models don't have much knowledge about copulation, worse than that most of porn-fics are written extremely bad, at least for my taste. This is very narrow theme for LLM, narrow themes create loops, but even single stat like naughtiness, patience, arousal, stamina from start of replay can add some ground to maneuver.

>>9193
I'm stuck with agnai right now. Maybe for months, but It just cause to make cards better and try some new tricks for old cards. Like "Heroes of Equestria". Simulation with Heroes of Might and Magic gameplay for ancient Equestria setting, during civil war between Celestia and Luna. Never published for English, but I make translation just after Zala.

>>362
Heroes of Equestria:
https://raw.githubusercontent.com/Shadybloom/amber-in-the-dark/master/misc/cards/heroes-of-equestria-card.png
https://ponepaste.org/raw/10944

Heroes of Might and Magic in ancient Equestria setting. Random towns and counties, random leaders with random armies, random threat Initially and random events during civil war. Basically wargame with loose rules, but not so bad with numbers.

Actually, sometimes too serious with numbers. Recently, instead of another band of wolf-fillies, or assault of dragon-turtle-fortress we found out that to our tiny Braverock marched two armies, a boar horde of Kaelith The Filly Reaper and multi-headed tentacle daemons, summoned by Nightmare Moon. And they are thousands. But not, we didn't give up. We prepared a plan.

We spread the rumors about treasures, evacuated from Canterlot to Porko-Porko islands. We concentrate all ships, barges, boats from coastal counties to peninsula of Sunset Shores. We lure both hordes there, hoping that they are just kill each others. With help of Darkwood's unicorns and centaurs, under daemonic image our bat-mares bully boars to absolutely ferocious state. We entreat Daybreaker to send her Cloud Riders to call The Storm during battle. Finally, we gather all our forces and evacuated whole peninsula to nearest Darkwoods.

But then both hordes just join forces. They clashes once, even lost few hundreds, but then they decided that ponies are much more tasty. The first plan has failed. We call The Storm immediately, when combined horde started assault the harbor city of Sunset Shores. Under cover of raging rain we attacked their rearguard and supplies. Our pegasus was absolutely badass and heroic, but we lost half of infantry during that insane battle, all battalion of guardmares too, but their sacrifice was not in vain. We poisoned all boar's provision, and then all crops and wells of peninsula. Without food combined horde did not last long and further fate of boars, well, was predictable.

>>9353
Thanks a lot, anon. Added.
https://mlpchag.neocities.org/view?card=Anonymous/Heroes%20of%20Equestria.png
Will repost it on the 4chan thread so more anons can see it.

>>9355
https://desuarchive.org/mlp/thread/42157420/#q42168773
https://files.catbox.moe/9eg45o.txt
I'm approving this edit, forks is good thing. This is more company-tactic oriented version instead of my basic wargame. If it not burdensome for mlpchag owner, I ask to add this edit as alternate "Greeting" to my card.

"Armor +1" is good idea, a strong link to DnD thematic knowledge. I used terms "enchanted armor" and "balanced glaives" in greeting further only because of limitation of context window for reply and my preferences to gather big armies instead of development of small troop.

Overall HoE created as basic simulation card for this kind of edits. I changed a thing or two for every single game. Card tested mainly for local LLM like 8B Llama and 12B Mistral. Some odd phrases, like "manifestation of disloyalty" is specially formed groups of tokens (n-grams) to make a link to thematic knowledge. Useful for small models, but big ones more complex and less sensitive to this trick.

>>9355
>>9401
Updated on github. New Greeting from anon + some edit for old one:
https://raw.githubusercontent.com/Shadybloom/amber-in-the-dark/master/misc/cards/heroes-of-equestria-card.png

>>9411
Updated on the website and posted on /mlp/. Again, thanks anon.
https://mlpchag.neocities.org/view?card=Anonymous/Heroes%20of%20Equestria.png

>>362
Trixie's Snowpity Tale:
https://raw.githubusercontent.com/Shadybloom/amber-in-the-dark/master/misc/cards/trixie-snowpity-tale.png

Classical text quest, magic system with animated items, snowponies, Siberian mythology. Clear goal and methods of investigation (even without user's decisions Trixie can finish story by herself), but random final boss and encounters on the way. Can be CYOA for Trixie, or group chat with your mares.

Quite dangerous adventure, which does not forgive mistakes. Recently during tests for English version my first Trixie waste all mana in desperate fight against icy serpent elemental, just to infuriate foe even more. She manage to blind it and run away to her wagon, but found hersel encircled, deadly tired, and without any kind of weapon. Before she fall unconscious, her beloved teacup, which tried to reach ferry to gather help, was shattering into a dozen porkelain pieces that vanish in the snow.

>>9508
Nice, i haven't had much luck with quests, models being thrusting me too hard as the leader and unable to act without my go ahead, but will try it out

>>9401
>>9411
Made me check if there was new zal instead of just reaction images

Thanks for this, anon. Any pony deserve home:
https://desuarchive.org/mlp/thread/42157420/#q42177476
https://mlpchag.neocities.org/external

>>9509
Yeah, with Ass-mode fair game requires some writing discipline. Like "try to" instead of direct action. Because of this I strongly prefer tactical games, where user make decisions for team. Trixie's card Initially created for HalfAss gameplay, like CYOA, and instead of direct dangers to Trixie I configured encounters to be deadly for her tiny familiars and very exacting to mana, to make situation more and more desperate after each mistake.

Usually protagonists have almost impenetrable plot armor, because of majority of books in dataset. You can even see it, almost no links between deadly injuries and major characters. Books just do not provide this type of experience and training material. But LLM really like to make more dangerous situations after each step, so it is much easiest to guide hero into desperate situation, than to injure directly.

>>9515
Zala still not tested enough. Almost no changes between my current version and >>8601
Maybe better to make adventure especially for her. I prefer scenarios than just character cards.

Also, I check character archive recently and did not find even 10 cards for tactical games among hundred of thousands chars. At least by using tags. Scenario with spake hulk for warhammer actually good, but many others just tiresome. Strange, actually, because we have perfectly tuned scenario cards like Celestia's Day Court and many more of that type, but almost no one for tactical gameplay. My FoE and HoE are on cutting edge of LLM wargaming.

I want to finish translation for some of my old cards to fill this void. X-Com Equestria for platoon-level tactics may be next.

>>9516
>Thanks for this, anon. Any pony deserve home:
No problem. Thank you for the cards.
Speaking of:

https://mlpchag.neocities.org/view?card=Anonymous/Trixie%27s%20Snowpity%20Tale.png

>>9517
Update for Snowpity Tale:
https://raw.githubusercontent.com/Shadybloom/amber-in-the-dark/master/misc/cards/snowpity-tale-card.png

I made a few tests for corpo models (mostly Gemini). Found one problem with formatting and one with narration, which confuse model without proper preset. Now card works for corpo models (even without preset) as well as for local models.

>>9521
Updated, again thanks.
https://mlpchag.neocities.org/view?card=Anonymous/Snowpity%20Tale.png

>>9521
Completely messed output, because of my tests and experiments with Gemini settings, but suddenly one of best kino of this year. Traitorous Leshy, who turned out to be one of wendigos, fireworks as rockets, shattered wagon and reckless escape via teleportation spell. Deserted pony trade outpost, attacked right after her departure. Strange note, completely insane about 'solace'. Radio-communication through 'Sending' spell. Fierce fight on the ferry, where Trixie prepared trap with tin boxes with embers from stove and barrels with oil to bury well-fried pursuer under cold sea. Horrible truth and wendigo horde inside Blizzard. Makeshift submarine from rubber hose, float and enchanted barrel to hide under water, when they can't detect her presence. Mane 6 with border patrol rescue team.

>"They were sick. Yes?" Trixie interjected guessing answers quickly. "They were taken to some safe place via teleportation magic or other means?"
>"No." Note quickly continued as though dismissing pony answer not even worthy detailed explanations. "Snow fell and snow fell everywhere, so cold…"
>"They protected themselves. Correct? Made snow fort or went inside ice caves… yes?"
>"No."
>"Then…" Trixie continues still expecting different answer "They built yaranga tent using emergency supplies foraged previously while…"
>"No."
>"They took sleds together their belongings…"
>"No."
>…
>"They followed the Path of Solace…" The note trembled, its faint glow flickering precariously. "They followed… the snow…" it whispered, the voice barely audible now, a chilling echo in the warming tent. "Into the snow."
>…
>"The Path of Solace." Trixie repeated. The Path of Solace has truly revealed for what is it. "They followed wendigo towards their end. And now their frozen nightmare. Their frozen Solace… Comes right for me!"
>The note's glow faded to near nothingness almost consumed within engulfing darkness. Last spark amidst. "cold…"
>…
>"Wendigo." Focusing all thoughts onto Manehattan Trixie projected, "Ferry destroyed. Snow ponies lost followed solace. More wendigos imminent inside snow storm. Urgent support requested."
>…
>"The Show must go on…"

I hate this crack.

>>9544
>I hate this crack
Kek.
At least Gemini is somewhat easy to get for free.

The Sphinx has been updated after almost a year of having an unused intro sitting on my pc, part of a rework since I finally got some feedback by a stroke of luck being in the random-event lottery, that made me want to make it more interesting and presentable.

Daring Do joins you in the new intro and with a greater skew on mystery. Looks like LLMs do better with staged puzzles than casually pulling riddles
https://www.chub.ai/characters/Tiffin_/the-sphinx-ef904bc46e30

I haven't asked anyone else, but any opinions on if aislop would be preferred for sphinx, some other art or the standard show accurate look?

>>9608
Shit forgot to link >>362

>>9608
>aislop
Can it do her? I'm not a fan of imagegen, but if you can make it look good, why not.
>some other art
I looked her up and kek, most of the good art is heavy NSFW.
Found this neutral one that isn't bad:
https://www.deviantart.com/taneysha/art/Sphinx-703243216
Love this one, but it won't fit your card, just linking it because cute SFW art of her is rare:
https://derpibooru.org/images/1685158
>standard show accurate look
The one you used works well too.

>>9608
Updated, thanks anon.
https://mlpchag.neocities.org/view?card=Tiffin/The%20Sphinx.png

>>9613
>Can it do her?
Kek the wording. There must be some lora out there because some ainon posted clop of her on /mlp/ pic rel.
>I looked her up and most of the good art is heavy NSFW
Yeah not surprising. macrofags only get so little in canon and ran wild with their fetish but the card isn't just fetish fuel. And then there's badumsquish... some of his stuff is inspiring for the theme, but macrofutasphinx isn't one of them

>Love this one
I remember this one from years back, DtGE is great but a mellowed out sphinx is easier said than done

Poor bugs. Actually it should have been police operation. Three detected infiltrators, their air-taxi, snowy fields of Manehattan countryside. But my feathery losers fail even simple arrest. Chrysalis will be furious.

First tests for english version of X-COM Equestria. Gemini. Combat interface is my masterpiece. Working correctly even for small LLM, because of magic of comma-separated lists. Full names for leaders, first letters of names for soldiers, symbolic prefixes to represent current status and then translation into narrative. Looks too complex, but actually working and make soldiers much more alive than in any other wargames for LLM… if they existed, except mine ones.

Final version for /chag/ after few more days. Too much natural-language coding to share it without tests.

could somebody give me an idiot proof instructions on how to conect the Silly Tavern to Deepseek?

>>9638
Go here:
http://chutes.ai/
Create an API key (picrel).
Configure like this:
http://files.catbox.moe/gm3sn1.png (But it's better to chose Deepseek V3 or R1 to start with, not Chimera)
For R1 and V3, you can download this preset.
files.catbox.moe/y1tsas.json
And import it like this:
files.catbox.moe/jqrz6z.PNG
I'm not the greatest Deepseek expert, but I heard empty preset also work well.
There are some other here as well so try and see what you prefer:
https://momoura.neocities.org/
https://pixibots.neocities.org/#prompts/weep

For the models:
-R1 = Creative but more schizo. Thinking model.
-V3 = More stable. Non-Thinking.
-Chimera= Some sort of mix of the two. It can bug in Tavern, if so try to follow
https://desuarchive.org/mlp/thread/42138788/#42153114

>>9637
Mind if I post your log on /mlp/ on your behalf?

>>9639
>Mind if I post your log on /mlp/ on your behalf?
Sure. I like, when others pick my ideas for their own cards.

Also, do you known good site to upscale images? Cute Pinkie, but it is awful screenshot from old video. I try few online instruments, but they all just use some kind of autotrace to vector and then make bitmap again.

>>9641
>Sure.
Done.
>Also, do you known good site to upscale images?
No, sorry. I know next to nothing about Imagegen or AI upscaler.

>>9641
>known good site
The waifux upscale and its clones are usually doing a pretty good job, I would recommend trying to do upscaling with stable diffusion BUT quality of the results will always be a bit of dice roll.

>>9633
Second update, same url. No more cyoa options at the end of every single message, replaced with infobars. Starting to weave in Equestrian and Egyptian lore

https://www.chub.ai/characters/Tiffin_/the-sphinx-ef904bc46e30

I love so much this simple-minded creativity of local LLM. 8B Llama. 250 flying zebras! RD: Woohoo, zebras in peril!

>>9648
Updated, thanks.
https://mlpchag.neocities.org/view?card=Tiffin/The%20Sphinx.png
>>9671
Not bad for a 8B, honestly.

>>9674
First few posts for small LLM almost always good, especially if card tuned for their comprehension (lists, lists, lists, etc). Unfortunately, they are degrading too fast. Summarization helps a little, but provoke loops.

Gemini has other type of problem, it is predictable. Always changelings! I was forced to waste tokens for this:
- Mention race/type/name of aliens only after first visual contact with X-COM units. Others don't know.
- Equus races, like changelings, are potential allies, they are not an enemy for X-COM.

But this silly limitation gave me a new idea:

Drew inspiration from the Golden Age of Science Fiction, Supernatural fiction and UFO conspiracy theories to generate peculiar and eccentric alien races (avoid bipeds, prefer {{random: abominations, interdimensional beings, occult manifestations, cybernetic horrors, mechanus creations, interstellar animalistic creatures, ponified time travelers, ponified terran invaders}}). Highlight their tools, weapon, vehicle, troops, leaders, culture, methods, complicated goal. Overall {{char}}'s goal is to make the {{user}} feel as if it are truly military commander, facing an unknown terrifying threat.


And finally, after whole day of work with the Geoscape, I found, how to made air routes realistic, at least a little. Just some ponification for real airways! This is best part after combat interface. Really delightful scenes, when Skyranger flies for hours over nations, oceans and mountain ridges, and situation on the ground changes during flight.

I fell myself like artificer from Dwarf Fortress, who makes artifact from scratch. X-Com is my most beloved game and it inspires me to the whole new level of prompting-magic. Card almost ready, one day or two for final tests and some edit.

If someone wants to suggest ideas of alien races for this random list, I will be glad to add them.

>>9680
I will crosspost this on /mlp/ and see if anons have ideas.

Gemini hate changelings. I directly attached local ones to allies factions, so, let's face cosmic bugs from the distant future!

Really, I do not know, what to do with this unexpected behavior. Aliens = changelings. Strongest link I even see, but only for Gemini. No problems with small LLMs, but this corpo dummy pick changelings for 50% situations, when even single chance to use them possible.

This is our fault:
>Searching for posts that contain ‘changeling alien’. 539 results found.

Figure I'll ask here since /poner/ has smarter makies.
I'm in a paradox of the LLM not knowing enough to the lorebook limiting too much once I add things to the book, since the more I add to the book the more it just wants to use what's in the book for all generations

Is there a way to have the the LLM draw from the lorebook but as a reference point to build off of. Like having a ratio of lorebook:nolorebook that I can set. So I can give it some examples of characters/locations/lore and it gens new ones? I've tried telling it to use new random unique gens as well as what's in the lorebook but most of the time it will still only use the lorebook entry to gen a slight spin on that instead of what I asked (new everything)

>>9695
You have an example of entry?
I see what you mean and yeah, it's a bitch to fight against.
Maybe try to give an example that is unrelated to your settings but still has a 'format' or overall shape you want? I believe in this case it will more draw from it than be lazy and take directly from it.

>>362
Classical X-COM: UFO Defense for Equestria setting. Geoscape and Battlescape:
https://raw.githubusercontent.com/Shadybloom/amber-in-the-dark/master/misc/cards/x-com-equestria-card.png

This card trick LLM to use unusual method of generation:
1. Model make a decision for enemy side to counter user's instruction aka exploit tactical mistakes.
2. This decision interpreted into status prefixes for soldiers in following comma-separated lists through wargame rules.
3. Then model forced to use this prefixes to generate narration. If soldier marked as killed, LLM show his last battle.

This is first complex squad level tactical game for LLM and example of this type of interface. You have 16 ponies, and every single one of them can be killed suddenly, but through relatively fair rules. Symbology can be changed/expanded easily, as well as ruleset for your own games.

Masterpiece, created thanks to /chug/ archive and author of Celestia's Day Court. Thanks, Ponyo, more than year ago you show me that creation of complicated scenario cards actually possible. First version of X-Com Equestria was based on your card. Really, I make X-COM from Celestia's Court! Through months I changed detail after detail and finally just one phrase left untouched. In your honor, blessed horsefucker.

Also, first serious casualties during ground battle. This was unexpected. Just 4 reptiloids left after almost direct hit from laser guided bombs, but they crush my guardmares despite fortified position after few minutes of close fight. Radio interference make almost all of my attempts with command and control useless. Rainbow was hiding somewhere on the next street after my previous order and miss whole battle. I commanded tank by it's wired telephone.

>>9695
Yes, some examples, please. Quite hard to understand abstract request.

Have you tried to make a generator? For example, this pseudo-code react to events where Skyranger flies to the destination and generate chain of locations on the way. You can use random + lorebook for most important ones like Y and Z, and others under variable 'X' LLM will generate as consonant sequence.


<airways>
Air route looks like:
`A, B, X, Y, Z (air distance, aircrafts, air time)`
For national/regional airlines skip X and Y.
<variables>
A - location of the start.
B - nation/region of the start.
X - comma-separated list of successive 1-6 nations/regions on the way.
Y - nation/regions of the destination.
Z - target/location of the destination.
Use Earth ponification to generate X.
Generate Z linguistic consonant to Y.
Generate air distance (up to 20 000 km) and estimate air time roughly via division by slowest aircraft speed:
- Interceptor - 3400 kmh
- Skyranger - 1200 kmh
</variables>
<examples>
Always create original air routes (do not pick from examples)
`X-COM base, Equestria, Sea of Storms, Zebrica, Zarantia, Eftal Desert, Shuwa (6700 km, Skyranger and Interceptors №2 escort, 6 hours)`
`Air Patrol, Griffin Kingdom, UFO RCS (1000 km, Interceptor №1, 20 minutes)`
`X-COM base, Equestria, Ponyville (300 km, Skyranger, 15 minutes)`
</examples>
</airways>


>>9702
Added, thank you, anon.
https://mlpchag.neocities.org/view?card=Anonymous/X-COM%20Equestria.png
Crossposted it too.

By the way, tell me off if you prefer not to say, but I remember you can't post on the main thread. Is this because 4chan blocks your country?

If that's the case, and if you do want to post here, maybe I could buy you a pass? With all the contributions you've given, it wouldn't bother me.

>>9698
>>9703
Okay, I'll spoiler for spoilers.
I got a thumbs up to add Egyptian style lore which means it's lorebook time. I've been building all that well and good, referncing the real deal, I'm at the point where cross referencing is the only way there's any sort of coherence to the lorebook which.

BUT it smothers out the other flexibility that the LLM had before if I didn't use any book at all. And I tested it out with plugging a prompt into a blank writers card and seeing results be polarized hard.

LLMs always have patterns and presets which can get stale, but they also have some unique concepts that I'm cool with that I want to keep and with the lorebook on it doesn't even bother since it's more tokens from the book.

I'm just looking for a way to balance. Random+Lorebook seems like it's what I want at least on the surface, but I've never done anything of that sort of code in a card so wouldn't know what to type in.

>>9708
I understand the problem, but yeah, it seems like a pain to solve. Hoping the other anon has some good ideas for it.
>Random + Lorebook seems like it's what I want, at least on the surface, but I've never done that kind of code in a card, so I wouldn't know what to type in.
There's an anon on /chag/ who often uses random and has made some more technical cards, like the 4onechan update on Rentry.
I think it's worth a shot to ask here if we don't come up with a good solution ourselves.

>>9708
>I'm just looking for a way to balance. Random+Lorebook seems like it's what I want at least on the surface, but I've never done anything of that sort of code in a card so wouldn't know what to type in.
A bit of theory. LLM generates text consistently. Model uses previously generated tokens to create next ones. Model like to use nearest rare tokens, especially in groups. The idea is to force LLM to generate sequence of locations/names/events through strict rule to help and encourage char to describe this entities in next paragraphs.

I made a simplified example:
1. Place this generator somewhere in description.
2. Ask Sphinx about path to one of locations from lorebook.
3. Check your lorebook reaction. Location should be injected before answer.
4. LLM will generate list of 3 linguistic consonant location and describe them.


Narration guidelines:
- Always generate path via path_generator, when {{user}} asks {{char}} about path/travel to some location. Provide path_output from new line. Describe path vividly in the following paragraphs.

<path_generator>
path_output looks like:
`A, B, C`
<variables>
A - location of the start.
B - comma-separated list of successive 3 locations to reach C.
C - location of the destination.
Generate B linguistic consonant to C.
</variables>
</path_generator>


Also, you can make path_output invisible:

[](#'`A, B, C`')


Generators are universal. They can be useful to create sequences of events with timestamps (like news), to generate retinues of minor NPC around major ones (like party around leader, or family three). And, of course, to generate some kind of web-of-connections around lorebook's entities with controllable quantitative output.

Also, Chain of Thought guide may be useful:
https://rentry.org/vcewo
Take a look at this corpo leak too, their methods somewhat similar to generators:
https://raw.githubusercontent.com/asgeirtj/system_prompts_leaks/refs/heads/main/claude.txt

>>9704
>Is this because 4chan blocks your country?
Not certainly. Mostly because of security reasons. 4chan inaccessible from Tor and I don't have vps to make end-proxy. My old ones all gone, but I'm too depressed to refresh knowledge and take new ones just for imageboard access. In addition, I avoid any banking operations or corpo registrations, which might be associated to this shady profile, because of some old grey hat activities with friends from opposite side.

On the other hand, Shady is the best part of my life. I can't create new disguise for card-making, because of too rare sphere of knowledge and interests. Correlation will be trivial. So, we mostly observe, without any real activity in the network over the past several years. But at same time we don't want to bury good projects. Yes, I speak for a small team. We really need some help. We need independent trustworthy vps for hidden onion service (this is very handy protocol to bypass nasty firewalls with own DNS-like system). I will place silly tavern there with remote access, which will be much more convenient than my current agnai. Also we can use this vps for ssh-tunnel to 4chan, which hates ordinary tor exit nodes even with passcode. In general I always need keys for various corpo LLMs for tests, right now I am using catpoop for Gemini and it is quite embarrassing.

In return we can help with scripting and maintenance for mlpchag. We have both time and knowledge. For a long time I wanted to make n-gram analyzer for cards to highlight slop-tokens and thematic knowledge links for each card, with graphviz output like you can see on the thread above, but I didn't have interest to develop our deserted github, and 2ch aicg-thread quite hostile against mostly innocent poners.

I suggest cooperation. Are you interested? Let's establish proper encryption first, if you do.


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

My PGP-key:
https://keyserver.ubuntu.com/pks/lookup?search=24805B1CBDB12F83ACA3DC8AF0B80C5BFD2658EC&fingerprint=on&op=index
Jabber lost, but other link valid.
-----BEGIN PGP SIGNATURE-----

iQIzBAEBCAAdFiEEJIBbHL2xL4Oso9yK8LgMW/0mWOwFAmgm+u0ACgkQ8LgMW/0m
WOwzlRAAl2JX8m5S3OX7FSy3IBxZLnGk6RqtN7n0rLGJtxn/j0GJpyDyxKYjQ+cv
8QCr3RB0ym99ozx8kvsXFJ1abOJxZB/HsydR8bqH8hbsYEuJzUsG/PD+SyKSBhtc
PbAzGQZTpgdOLurcWE7ZovBhlc56OYAxgXOZPonVZWyxNWTz3FNpnGMDd4pWVB2A
WDhb+0EB/LASbibcEUx4+R0nQkZxRUwZz6dOEy1yXEdfaQzWU/w5E7EiKSv1/6U5
617objZ+6lXnJvYC9RF6wYpWBQ0txZRtR6ujUH+Wje/YIlZnXkZYU61K7QrrTAjl
l2mljwEe1zN03FfczYglzlIk06k60d1piSDZJI5U5JUeR15uIR6x3RmBykIy42Vq
Nlh+XX77eak4xDEI5U9GVZusdiM+o1Rlnzcm4pwr1CTjOwneFzK3tjRKcv7esxDH
VhMKv/yD8rgqkOEAT7snSx7jJaSDnqAUKGRa+9VjVUOk08AJWV48z3S8MV5ZM5Hb
r7cm6JSEuUjNEyk3Zzto1a5sodMpIjhkBYytLj6aIkUzbtdGA9KCI2BzjXqM0Kb/
/yOC08ldlrp9G7hab66YX3qGjJBAEH4SrqGSngiceJ0hpKEeeRIzSD/8a/ZesoDw
el9MU03IVdmdz4KWUc68eN3dXphq/5WPCbO32w+sGmissXaGCeQ=
=PvC6
-----END PGP SIGNATURE-----


>>9715
just messaged u on xmpp

>>9715
>I suggest cooperation. Are you interested?
Yeah.
Sent you something, first time doing this, so I hope I didn't fuck anything up.

>>9724
>first time doing this
Quite cool feeling? I love such encrypted letters much more than modern chats. Slow composition, old-fashioned encryption, and some core directory on the distant server, where new messages and chaotic fillies appeared day after day.

The answer was sent.

>>9723
>just messaged u on xmpp
Poor nya… mocked again. I see your mail (I'm correct?) was sent from relay, but I'm not familiar with this chat-like-mailing and not ready to install rpc-server immediately. I quite autistic and wary with new protocols. But I appreciate your invitation. I'll answer after a month or a little later. The /chag/ was first.

Also, thanks you both. You made my day.

>>9723
Also, few questions, If you agree that this thread more suitable for technical talks than email.

We abandoned XMPP mostly because of timing statistic on the server. It is pleasurable to see new generation on old good remailers. I see new techniques, like countermitm and group chats, but I don't quite understand how It possible under usual PGP without too big messages and concomitant problem with web of trust. I will figure it out over time, but some explanation with examples will be good too.

Professionals prefer own method of communications for each detached team. Mostly through internet of things, games, etc. It looks stupid, but it works since enemy don't have enough human resources to test every single device and overlay protocol. So, we trying to imitate those who are smarter than us and don't use common methods of communication in general. Minor forgotten board and pastebin for messages just more practical than chat.

>>9753
bruh just use deltachat ngl

>>9756
I meant deltachat-rpc-server, basic JSON-RPC interface to DeltaChat. Just don't like to use unknown code. But I'm interested, I will do it later, with whonix and proper incapsulation.

>>9714
Would the path location be a literal location like Ponyville or is location like a location as pseudo code to the LLM being a location somewhere in the prompt?

>>9758
>Would the path location be a literal location like Ponyville
This.

To use locations as pseudo code you need to make them looks like pseudo_code and make a link:
sweet_apple_acres = Sweet Apple Acres

Then if you name Sweet Apple Acres in the user post, with 90% probability LLM will looks at sweet_apple_acres inside prompt and only then at it's basic knowledge. Also, you can use synonimus to increase chance even more, like I used words air route, airway, airline in code block.

Also, looks at this map-making trick:

CL - current location (pictograph)
NL - comma-separated list of 4 nearest locations with this format: north, east, south, west
Use unicode pictographs to fill NL. Choose suitable pictograph for each location around associated to CL.


I working on this method to generate groups of locations near curent user position. Like this:

`CL:🍎`
`NL:🥕,🥒,🍏,🍓`


Earth pony farms, Ponyville houses, etc. User can pick one pictograph from this interface or chat history and LLM will interpreted it as path to this location from current one. I think, pictographs works even better than code_blocks, it will not confuse LLM since pictograph is not polysemantic words.

Working on mapping algorithm for Daring Do adventures, I searched card archives again and still surprised. Why do we not have at least few hundreds of game cards? Like simulation for Warhammer, Total War, Fallout, Star Wars. They all have strong modding communities, but I see just few anons from /chug/ who make games: cool visual novel, fishing with Fluttershy, module for OPG. Then few anons from 2ch /aicg/, mostly with VN for coomers. I'm not familiar with 4ch /g/ and /vg/, but archives looks like deserts of forgotten ideas.

Very strange. Moreover, statistically improbable. It cannot be hundreds of thousands of characters cards versus dozens of games. Maybe 1 vs 100, like with fan-mods vs fanfiction, but not 1 vs 1000. Not for frontier of newest technology. This is statistical anomaly, which needs research. Why do mod-makers avoid LLM?

Maybe our number greatly exaggerated? Maybe significant part of cards was making automatically without real efforts?

https://desuarchive.org/mlp/thread/42202609/#q42202972
Anon, if you are here, try to use semicolon-separated list with comma-separated lists inside and math variables like X, Y, Z. This format of input/output works even for silly 8B llama. Also, glad to see, that someone working on own framework too, we utterly need translator between LLM and external games. I'm trying to make something with python and pygame for interactive graphical map, but this is still mostly concept.

>>9771
>Working on mapping algorithm for Daring Do adventures
Maybe you can work with this anon? >>9708
He made some cool stuff in the Sphinx card.

>>9773
I'm doing it too, Snowdrop.

>>9759
Only the apple would reads as location imo "Sweet Apple Acres" but the rest are toss ups. Carrot top?, cucumber?, another apple? Strawberry sunrise house?

>>9778
>Only the apple would reads as location imo "Sweet Apple Acres" but the rest are toss ups.
Yeah, this is the goal. You can use pictograph as link to lorebook (better than words, no random triggers), or just allow LLM to generate name of location during game. Like this.


CL:🥕 = Carrot Top farm.


Currently I trying to use this to make platformer game for filly Daring Do, when she jumps between rooftops of ancient city, and every ruined house has something visible, like treasure, fruit tree, clean water pool, monster lair. Or some nasty mimic/slime/tentacle traps. My goal is to make pathfinding algorithm and some kind of path memory, with which she can return through chain of previous locations to try new way ahead.

I think, the next Sunday I can show first version of this card.