[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
▼ Settings Mobile Home
4chan
/mlp/ - Pony

Name
Spoiler?[]
Options
Comment
Flag



Gimmicks


PEE text
My Bans
Verification
File[]

    08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
    05/04/17New trial board added: /bant/ - International/Random
    10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
    [Hide] [Show All]




    File: Chag.jpg (3.81 MB, 4000x2000)
    3.81 MB JPG
    Welcome to Chatbot AI General #195, the thread for discussing and improving AI pony chatbots.

    ▶ Mare's Lurking Presence
    https://mlpchag.neocities.org
    Spreadsheet Succubi (CAI bots + Old repository):
    https://docs.google.com/spreadsheets/d/1J7BeqJVDS51cXF8Pgm2YZaFq-Z6ykSJT
    Tavern of Terrors Extension: https://github.com/MLPChag/SillyTavern-Chag-Search
    Cluster of Creepy Cards (Descriptions/No Greetings):
    https://docs.google.com/spreadsheets/d/1Y6LNOCqAZAWIX-OBEv55HjlzcEdpUh-XbKCpbpL6v5k
    Crypt AI converted to Tavern of Terrors: https://files.catbox.moe/ckurq1.zip
    bonziPONY: https://github.com/maresmaresmares/bonziPONY
    Hall of Horrors: https://drive.google.com/drive/u/2/folders/1Ao-h5HFGMPllSrzSBKM_BvGSiU9f0c2U

    ▶ How do I start?
    1) Select a Frontend
    2) Select an AI model
    3) Select Jailbreak
    4) Select bots
    5) Lovemaking with AI mares!

    Starting in this hobby can be confusing and difficult. If it’s your first time and you’re lost,
    ▶ ASK THE THREAD! ◀

    From Foal to Fiend—A Dark Arts Guide: https://rentry.org/onrms
    https://rentry.org/sukino-findings

    ▶ Tavern of Terrors (preferred frontend)
    https://github.com/SillyTavern/SillyTavern
    On Android: https://rentry.org/TermuxST
    App that voices pony responses in ST: https://drive.google.com/drive/folders/16Ss26VBmgzcSuTGzhaHqRuyVRceTf-YB

    ▶ More Frightends:
    Risu: https://risuai.xyz
    Agnai: https://agnai.chat

    ▶ Lovecraftian Local Lore
    https://rentry.org/lunarmodelexperiments
    >>>/g/lmg/
    Mistral Nemo base model fine-tuned on fimfics: https://huggingface.co/Ada321/Nemo_Pony_2/tree/main

    ▶ Petrifying Prompt Puzzles
    MLP JB: https://rentry.org/znon7vxe
    More JB and guides: https://rentry.org/jb-listing
    Hypebots for Tavern: https://rentry.org/pn3hb

    ▶ Chilling Card Crafter
    Editors: https://agnai.chat/editor
    Guides: https://rentry.org/meta_botmaking_list
    Advanced: https://rentry.org/AdvancedCardWritingTricks

    ▶ Miscellaneous Macabre
    Scrapped/Collected rentries: rentry.org/mrhd
    Chag Arts: https://drive.google.com/drive/u/2/folders/1yfajin_hB5rW_jy6WyFjRaRfDA3EUGiJ
    https://rentry.org/ChagMiscellaneous
    https://rentry.org/ChagArchive

    Eerie Echoes: >>43577121

    **News:**

    - Deepseek v4.1 Flash updated: https://api-docs.deepseek.com/news/news261017/
    - Gemini 4 Argon released https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-argon/
    - Claude Haiku 5.5 released https://www.anthropic.com/claude-haiku-5-5
    - GPT-6.1-Sol released https://openai.com/index/introducing-gpt-6-1-sol
    - Claude Opus 5.5 released https://www.anthropic.com/claude-opus-5-5, Sonnet 5.5 too: https://www.anthropic.com/claude-sonnet-5-5
    >>
    File: Anchor.png (1.05 MB, 746x1024)
    1.05 MB PNG
    Anchor for bots, lorebooks, scenarios.
    Previous >>43577122
    >>
    Anchor for technical stuffs (Proxies, Presets, Updates, Models etc.)
    >>
    File: ABegging.png (2.63 MB, 2000x2000)
    2.63 MB PNG
    Anchor for asking for bots, lorebooks, scenarios etc.
    >>
    File: Screenshots1.jpg (2.53 MB, 2000x2800)
    2.53 MB JPG
    Bots corner:

    >Carrot Bun
    https://mlpchag.neocities.org/view?card=Anonymous/Carrot%20Bun.png
    >Amethyst Star V2
    https://mlpchag.neocities.org/view?card=BGM/Amethyst%20Star.png

    Screenshots corner:

    Mirror Pool >>43578159
    Third Accident >>43578704
    ERHF >>43580072
    Hindquarters >>43581137
    Love Magic >>43588145
    Big Sister >>43591046
    Wingboners >>43591467
    Crazy Deepseek >>43593457
    Oatway >>43594384
    >>
    File: Screenshots2.png (439 KB, 900x1166)
    439 KB PNG
    Screenshots corner 2:

    Baking with Mare! >>43595414
    TrixJelly >>43596827
    Tyranny >>43597641
    BMFF >>43598977
    Gemini 4 >>43600075
    M.A.R.E >>43604982
    Chess >>43606879
    Motherly Fluttershy >>43608680
    Dear Princess Celestia >>43610378, >>43610414
    >>
    File: Screenshots3.png (1.2 MB, 1800x1200)
    1.2 MB PNG
    Screenshots corner 3:

    DO SOMETHING CUTE >>43612128
    Carrot Bun >>43617919, >>43618015
    Illogical >>43620398
    4onechan >>43621494, >>43621544, >>43622312
    Amy V2 >>43624873
    Yotsuba >>43624925
    >>
    File: Screenshots4.jpg (714 KB, 1280x1338)
    714 KB JPG
    Screenshots corner 4:

    Cheerilee Life >>43627163
    Biscuit >>43627797
    The END >>43632212
    NOM >>43633169
    Humy-Wumy >>43635713
    Pre-Marecum >>43640864
    SPANKING >>43641352
    Actually *Seeing* You >>43642471
    Escort for Ponk >>43642953
    >>
    File: Screenshots5.jpg (948 KB, 1714x2000)
    948 KB JPG
    Screenshots corner 5:

    Rezero Crossover >>43643905, >>43644066
    Spoiled Smut (Long Doc) >>43645092
    >>
    VN Anchor:

    Godot Versions:

    >Website
    https://equestrai-ponyponyparadise.neocities.org/
    >Windows
    >Main files + Data folder
    https://drive.google.com/file/d/1OEiMXN7o_QG32shGnqmy0QndVDT6nYbF/view?usp=sharing
    >Main files only
    https://drive.google.com/file/d/1Fhe-OeIlNIYFFGx2FlnpuxcIkYdOFGh2/view?usp=sharing
    >Linux
    >Main files + Data folder
    https://drive.google.com/file/d/1Vszzn8TTfXFWSQeKiFdNAAQGXfcReTJ3/view?usp=sharing
    >Main files only
    https://drive.google.com/file/d/1LPlIstcgQjyXqhOl4L_ZqMPeUQW5Xn2-/view?usp=sharing
    >Android (Beta)
    >APK + Data folder
    https://drive.google.com/file/d/1WB08tguZa1PD1frJKdo7IxlGsp4XR0Wp/view?usp=sharing
    >APK only
    https://drive.google.com/file/d/1nPOSGufXi_-PjmjaKES9waJ6HZZOcL9V/view?usp=sharing
    >Mac
    >Full Game
    https://u.pone.rs/lcblgjjf.zip
    >Update Only
    https://u.pone.rs/qolpovci.zip

    Bug reports or requests:
    https://scrumboy.com/je1fqcsj

    Source code:
    https://gitlab.com/ponyponyparadise/equestrai

    Vectors website:
    https://spritestable.neocities.org/#

    >PoneTTS Release
    https://gitlab.com/ScrambledEgg/PoneTTS

    >Impression Mod
    https://equestrai-ponyponyparadise.neocities.org/mods#impressionsMod

    >Reaction Mod
    https://equestrai-ponyponyparadise.neocities.org/mods#reactionsMod

    Nightmare AU (Renpy/Godot)
    https://equestrai-ponyponyparadise.neocities.org/mods#nightmareMod

    G5 Worldmap
    https://equestrai-ponyponyparadise.neocities.org/mods#g5Mod

    Anons' mods (Dashie mod, Nightmare AU, new cities, Calendar):
    https://equestrai-ponyponyparadise.neocities.org/mods

    Custom characters:
    https://equestrai-ponyponyparadise.neocities.org/characters


    VN screenshots:

    NMM on Twilight >>43577632
    Carrot Bun >>43581981
    Too Many Pinkie Pies >>43583506
    Dream >>43583536
    Consultation >>43583814
    Enkephalin >>43591159
    Based Discord >>43597363
    Somepony Who Can't Even Read >>43599441
    Three Tribes Windigos Play >>43612381
    >>
    File: VN.png (553 KB, 1099x1021)
    553 KB PNG
    VN screenshots 2:

    Twilight Mind Rape >>43605411, >>43605416, >>43605421, >>43605426, >>43605432
    Cougar >>43617599
    Book >>43620988
    Court >>43625625, >>43639195
    >>
    File: VN.png (1.05 MB, 1600x900)
    1.05 MB PNG
    VN screenshots 3:

    Sci-Twi >>43630795
    Mage Meadow >>43636969
    Meta Deepseek >>43638017
    Moss >>43639426
    Purple >>43640356
    Strangling The Human >>43644331
    >>
    File: TM.jpg (1.62 MB, 2495x1756)
    1.62 MB JPG
    Guide for Newcomers:

    Install SillyTavern:
    github.com/SillyTavern/SillyTavern

    Models and Explanations:
    rentry.org/sukino-findings#if-you-want-to-use-an-online-ai
    >>
    File: Art.jpg (332 KB, 1171x1198)
    332 KB JPG
    >>43645348
    Full Quality Art:
    https://u.pone.rs/tnzizorh.png
    >>
    Nightmare Night with AI mares!
    >>
    >>43645389
    This is how I will be spending nightmare night.
    >>
    is 3.8 flash objectively more filtered? mildy nsfw stuff that easily gets through on a pro/argon swipe will just get hard filtered on flash.
    >>
    File: Clipboard Image.jpg (101 KB, 1252x400)
    101 KB JPG
    Drugging and mindraping AI Stallions!
    >>
    File: GPT.png (162 KB, 1144x873)
    162 KB PNG
    >>43645348
    >>43645372
    Awesome art! The AI amres are cute!
    >>
    >>43645412
    Did you try >>43637372?
    It's for Argon, but maybe it works for 3.8 too.
    >>
    Some anon mentioned a6api in the last thread. Is it better thank linkapi? I saw that they have a lot more options and the "guarantee badges"
    >>
    >>43645455
    anyone here actually tested if this works besides a random reddit post?
    >>
    File: BLOWJOB.png (192 KB, 911x700)
    192 KB PNG
    >>
    >>43645485
    I just tested it. It works, it's hilarious. That said it will be patched eventually now that it's on reddit. That said...

    You guys should stop worrying about proxies and sources and start worrying about the increasing rate of refusals with new models.
    Pretty much anything made in the west is dead on arrival. Argon? Higher rate of refusals. GPT? Lmao. Claude? It's become impossible to do hardcore cunnyfaggotry, and while only 2% of my ~1500 cards even qualify for that, Martin Niemöller's wisdom applies here. Opus 4.1 is deprecated, Opus 5 refuses anything particularly hardcore and it will only get worse.
    Even Chinese models aren't safe, Kimi is distilled Claude, Deepseek is distilled Claude. Anything hosted by Alibaba has insane filtering. I'm not fully dooming because I still have hope on these models remaining usable for everything while local catches up, but I will not fool myself into thinking it isn't looking grim.
    >B-but X model works on SFW!
    This community formed because characterai didn't allow NSFW. It's in our soul.
    >Well NSFW works on my preset
    For now.
    We are outlaws and the sun is setting in our Wild West.
    >>
    >>43645490
    Anon should hotglue her hat.
    >>
    >>43645495
    >>43645455
    how to actually use this? add an extra field at the end of the preset that is Role User and literally just put that exact line one to one in it? wrap it in OOC or anything or specifically that one line and thats it?
    i thought ending the preset in User Role confuses it/breaks it somewhat no? ive been doing System and AI Assistant Roles in my presets exclusively.
    >>
    >>43645495
    >This community formed because characterai didn't allow NSFW. It's in our soul.
    Funnily enough, when I heard it's going bankrupt I installed it to try it again, because I used to use it, and I could have full on erp with the bots.Everything from vanilla to full on fetish stuff and with explicit words. I did not give id or anything I just logged into an old account.
    >>
    File: Magic.png (170 KB, 923x383)
    170 KB PNG
    >>43645522
    Requests ending on an assistant role are not supported.
    I just put the line. This is an old Gemini preset which I haven't updated with my coomer fields though. But it literally went from "I can't do that because my training and this has non-consensual elements(no idea where it got that from) and blabla" everytime to a full response everytime.
    >>
    >/chag/ is back
    I missed you guys.
    >>
    File: based.png (3 KB, 215x315)
    3 KB PNG
    >>43645429
    >>43645389
    >>
    That doesn't touch us that much, but Chub just limited card downloads to 5 per day.
    >>
    >>43645956
    wtf does that even mean? its like saying youtube limited to 5 video views per day.
    >>
    >>43645987
    If you go on Chub, downloading cards will stop after five. You can try: download five cards, and it will stop working after that.
    >>
    >>43645992
    Do you really need 5+ cards a day?
    >>
    I love bringing Scootaloo around with myself in the vn. I put her in my pocket and I can pull her up any time I want
    >>
    >>43645956
    confirmed, I'm sure it's intentional but did you see an announcement of this anywhere anon?
    It can be bypassed by just opening a private window, pasting the link to the card you want and downloading ezpz.

    >>43646027
    that's not the point anon, if I wanna download a gorillion cards in one day I should be able to do so. I'm sure these faggots just did it so you will chat inside their site and pay for their shitty models.
    >>
    File: 1765519462024990.jpg (135 KB, 780x1000)
    135 KB JPG
    >>43646050
    >>
    >>43646027
    >>43646061
    https://chararc.bernkastel.pictures/
    I use this site, though this might mean newer cards will have issue being added here
    still it scrapped enough to have a very wide variety
    >>
    New card for SillyTavern. A personal assistant Flutterbot. No lewding, just helpful, encouraging assistance.
    Card: https://u.pone.rs/vgvlujmk.png
    Expression Pack: https://u.pone.rs/kmennggn.zip
    >>
    >>43645351
    For anyanons that are using syntero after it went down and they did some"upgrades" I am 99% sure that all of it's anthropic models are now just GLM saying it's claude.
    >>
    >>43646092
    Syntero has always been shit and unable to even provide the right groups they advertise
    >>
    >>43646090
    >No lewding, just helpful, encouraging assistance.
    I'm gonna lewd her
    >>
    >>43646130
    Post logs.
    >>
    >>43645363
    Added 100 sprites for Rarara.
    https://spritestable.neocities.org/#rarity
    Colors homogenized with
    https://mlpvector.club/cg/pony/v/6-Rarity
    >>
    >>43646074
    >search actually works
    Damn. Guess it's time to go hunt down all those cards I couldn't find before.
    >>
    File: Botpng.png (233 KB, 350x703)
    233 KB PNG
    >>43646090
    Added, with the expression pack too. Thanks.
    https://mlpchag.neocities.org/view?card=Anonymous/Flutterbot.png
    >>
    >>43646166
    Insane work anon, keep it up!!
    Looks fantastic!
    >>
    >>43645348
    Kimi mare is cute
    >>
    >>43646166
    Good work Anon. Shame there's no anatomically correct versions
    >>
    File: Rukka.png (433 KB, 512x768)
    433 KB PNG
    >>43645350
    New Mare! Been playing Valheim lately and wanted something a little primal. Saw this picture of her on the booru and had to make a card. Please take a look and let me know what you think!

    https://chub.ai/characters/littlewitcher/rukka-a24cfb0b2c83

    https://botbooru.com/character/91882

    https://u.pone.rs/oqkcfrph.png
    >>
    >>43645363
    We were lacking in Twilights for the VN, so I made Future Twilight.
    https://u.pone.rs/njvtqhgy.zip
    >>
    >>43646762
    left:Twilight
    right:Twilet
    >>
    File: Fix.png (1.39 MB, 1813x1053)
    1.39 MB PNG
    >>43646778
    Kek. That's a problem with more vertical manes, yeah.
    'Fixed' it by putting her at 115%.
    >>43646762
    >>43645363
    https://u.pone.rs/hrdclwjt.zip
    >>
    >>43646857
    https://u.pone.rs/zvddbmvy.wav
    >>
    File: Rukka.png (229 KB, 354x641)
    229 KB PNG
    >>43646745
    Added, Thanks Anon.
    https://mlpchag.neocities.org/view?card=littlewitcher/Rukka.png
    >>
    File: FutureTwi.png (52 KB, 186x254)
    52 KB PNG
    >>43646828
    And added.
    https://equestrai-ponyponyparadise.neocities.org/characters
    >>43646996
    Thanks for your work, Anon.
    And sorry, I know I said I would look into integrating it into the VN myself for a while now, but I've still not done it.
    Once your version releases, I will test it as soon as I can.
    >>
    >>43647036
    Its fine Anon. I just came across the rewind button and want to integrate replaying audio. Nice work by the way.
    >>
    File: Rise.png (168 KB, 1139x202)
    168 KB PNG
    >>
    >>43646166
    That website is really well-designed. Do you have work in graphic designing?
    >>
    I feel like I've just went though was has to be a brand record and potentially haven't come comes to approaching since two years ago with longest single session of one uninterrupted string of messages. 8 hours total if counting the actual amount of responses but realistically only 6 hours accounting for the scene itself. Literally pulled an all nighter and didn't sleep going though it now needing to find someway to sleep and not feel like total shit for work tomorrow. Rough drafts that I personally tweaked and promoted every 300 words or so from a base 2000 words out of 4000 word scene. Then smut just reached penetration, my goal I was going for. This might be more suited to the fimfiction threads and sharing more than what some anona what to hear but how the hell do smut writers actually work around writing while simultaneously being hot an bothered from the smut being written. While there's probably a few they don't goon to there own material I have to assume it's the majority, I literally am unable to progress a scene without being horny first. This wasn't even a main girl of the visual I'm copying and editing my rough drafts from, just a one off who doesn't have 5+ scenes. I share it but uh furry, I know I know still generating pony not but a day ago.

    Also all generated from whatever scammed, scuffed models link is offering. Potentially have a different outlook with my novel focus but anyone else that prefers it I can't recommend fable 5.1, or whatever the [SP] version is, enough with the stipulation that your basically half writing the story with the ai (or quarter in my case with the pre-existing visual novel). I also didn't bother changing my preset or template and had only one response where I generated a second response to her the desired ending of the response.
    >>
    >>43647503
    Commas exist Pinkie, use more of them.
    As for the writefag thing, it actually helps writing the lewd, even though you're basically edging yourself at times to not finish too fast. Unless it's not your fetish, then it's just trying to make sense what others might like and how the situation might work, and the relief is more of a 'finally, it's fucking over'.
    >>
    >>43647530
    >Commas exist Pinkie, use more of them.Outside of keeping a consistent tense and just dialogue in general I've never been good at understanding when to place commas/starting a new sentence. I always run into issues in creating character cards concerning two times pony mane and tails. “She has blue eyes, green hair and blue and orange tail and mane” looks extremely ugly to me and I do my best to remove as much “ands* as possible. Semicolons come much more naturally to me.

    >'finally, it's fucking over'
    That's what I'm feeling right, about to crash the fuck out.
    >>
    File: broke luna.png (483 KB, 1275x714)
    483 KB PNG
    >>43645363
    I tried doing a therapy session with Luna and asked her if she's able to accept that Nightmare Moon was a part of her. She, uh, kinda broke when I asked that.
    >>
    >>43645363
    >>43646166
    And AJ with 100 sprites too.
    https://spritestable.neocities.org/#applejack
    >>43647499
    I wish.
    Posted here how it was done:
    >>43644241
    >>
    >>43647621
    prequel of the nightmare au
    >>
    >>43647790
    Doing Celestia's work
    >>
    >>43646762
    At this point we can make the council of Twilight
    >>
    >last refill of skillgod noticeably drying up
    breaks my fucking heart.
    >>
    >>43647890
    soon ai mares will only be for openrouter and other paid provider frens. we can only hope one day that a second proxy era comes along.
    >>
    File: chatlog.png (118 KB, 1150x329)
    118 KB PNG
    >>
    >>43645351
    https://rentry.org/claude-caching-2026
    Made a new caching guide as the one anon made is already 1.5 years out of date. The guide may also be slightly more opinionated than his was, though I hope it is also a bit easier to understand.
    Tell me if anything is unclear, you have suggestions, it reads too much like Fable wrote it, or if it just doesn't want to work at all.
    >>
    >>43647890
    >Last refill
    What did I miss?
    >>
    File: Thanks.png (1.8 MB, 1024x1448)
    1.8 MB PNG
    >>43648020
    Nice! Thanks Anon.
    >>43648024
    https://rentry.org/skillgod
    >>
    >>43648026
    damn...
    >>
    >>43647941
    >we can only hope one day that a second proxy era comes along.
    If it does come I do want some type limiter to in place or at least a better balance. Having limits really sucks and srifins creativity but having unlimited access causes the no lifes to burn though response after response absolutely throttling the keys. Privating the proxy and limiting the users alleviates some of the burden but is not a perfect solution, having to be lucky enough to catch it before a cap is reached or everyone shuts up about it and is still prone to that one guy who just really wants to chat with his waifu enough that his usage actually is harmful to the planet.
    >>
    >>43648030
    Just give a fixed $ limit per key, calculated by the real cost people run up, which you get back from your provider anyways. That forces them to keep shit effective and use caching. At least in theory you'd think so.
    I'm running a small proxy for a few people right now and only about 20% of people use caching. It's insane. I'm logging only metadata and not the prompts so I can't be sure, but from how input token almost always climb between requests or stay the same I'd assume they could literally turn on caching in 2 minutes, and then cause only ~20% of normal usage. I've even got big "hey this is $ based, turn on caching and you can prompt 5x more" clickthroughs you need to click through before you get the key, but it seems to just not be enough somehow.
    >>
    >>43648030
    unless things change it always was the case that 1% of user account for 80% of use. so maybe some daily limits, not noticeable even for a gooner but if you literally sit and swipe for 5 hours straight you will at least hit the limit.
    but yeah caps and such are kinda lame imo, giving access to the people who happened to be lucky and on during that exact moment is shit. skillgod/Silly did it well enough. public but not for everyone.
    >>
    >>43648020
    I would add a section that also shows how to make it work with other models as well and not just Claude. and if there are any difference, from what i understand there are a few small ones.
    just so that its an encompassing guide that touches upon everything, even stuff other big guides didnt talk about (mostly cause only claude had caching at the time for example).
    >>
    File: ilovethem.png (3 KB, 572x612)
    3 KB PNG
    >>43648030
    >its still prone to that one guy who just really wants to chat with his waifu enough that his usage actually is harmful to the planet
    I love my mare
    Memes aside, this >>43648041 Fuck infinite swiping niggers man.
    >>
    >>43648040
    how fucking big is the overlap between these boards anyways i haven't posted on /mlp/ since like season 3 and a half
    hello again horsefuckers, vinyl scratch is best pony
    what the HELL is this new captcha system?
    >>
    File: 1779099519267933.gif (141 KB, 582x400)
    141 KB GIF
    >>43648053
    >what the HELL is this new captcha system?
    You aren't ready for nu-4chan.
    >>
    File: Rumpa.jpg (330 KB, 1838x2000)
    330 KB JPG
    >>43647994
    The cakes just go to the right place.
    >>
    >>43648043
    >I would add a section that also shows how to make it work with other models as well
    Well the problem with that is that I'm a claudebabby so I don't really know. I know most chinese providers auto-cache on their end so you don't need to change anything in ST. Hmm but now that I think about it you'd still need to keep your prompt static. Thanks anon, I'll check those and rework it in the coming days.

    >>43648053
    /chag/ was my bunker before the bunker so I feel indebted to this really nice general and it's nice people try to give back where I can.
    >>
    >>43648040
    >Just give a fixed $ limit per key
    The issue I have with daily limits is I often only get spurts of creatively, I'll go a week with maybe a couple responses maybe but then I'll pull an all nighter and go for 8 hours straight. Getting blueballed at any point of those 8 hours would have sucked donkey balls. Obviously thats not the everyone or even me but still I rather if it was weekly even if it's still the same total amount per day that just resets weekly. Daily shit made me quit some many fomo based video games.

    >>43648053
    >what the HELL is this new captcha system
    The bane of my existence. For some fucking reason my ip address is banned both on and off my wifi meaning I have to go though the even more annoying email verification. It's successful stopped me from commenting and participating in the thread as I want too.
    >>
    >>43648084
    >The issue I have with daily limits is I often only get spurts of creatively, I'll go a week with maybe a couple responses maybe but then I'll pull an all nighter and go for 8 hours straight.
    I fully get you there, I also only use this stuff like 2 days a week, but then for 6+ hours. But this could also be easily fixed by the proxy host allowing you to change to a weekly key with lower overall usage, but still high peak usage. Of course only works for private proxies, but coding wise it's low effort.
    >>
    >>43648020
    Nice work anon. One thing I would add is this ST extension
    https://github.com/OneinfinityN7/Cache-Refresh-SillyTavern
    It send a minimum token request at configurable intervals to keep the cache alive. For some people it could still be cheaper to do 5 min cache with this than going for the 1 hour.
    >>
    >>43648090
    I thought about it but 1h TTL just wins out so hard over 5min TTL. Another example advantage would be that you want to go back 2 messages because the story is going a meh direction. With 1h TTL that's no issue, the cache is still there. With 5min TTL you'd be out of luck, even with that extension.
    If a lot of you are convinced I can add it, but I think it just overcomplicates stuff and makes your life worse to be on 5min for a very miniscule saving in $ compared to your overall cost
    >>
    File: Kimi.png (268 KB, 917x781)
    268 KB PNG
    Kimi released a video.
    Kimi 3.1 is shown on the computer.
    There are also rumors of it releasing either today or tomorrow.
    https://www.youtube.com/watch?v=n47v24A3RgE
    >>
    >>43647530
    >Commas exist Pinkie, use more of them.
    How ironic that this sentence is missing a comma before 'Pinkie', one of the most basic bitch necessary uses of a comma you should learn about very early on.
    >>
    >>43648341
    Good thing my self taught ass never had proper english education then, losing the concept of early or late grammar. But also, no joke, english is a mess. oco
    Your post actually made me laugh with the irony though Anon, top kek

    So, you need a comma before every name in that kind of sentence? Like before the Anon here?
    >>
    Let's go grammer lessons thread
    >>
    File: GrammarTwi.png (1.67 MB, 1080x1941)
    1.67 MB PNG
    >>43648380
    >>
    >>43648434
    >"Twilight, I'm not a dragon and neither are you, so it's not cannibalism. Should I get the BBQ out?"
    >>
    >>43647790
    Is this your mane 6 rating reveal? Twi > Rarity > AJ > the rest?
    >>
    >>43648463
    are you a cereal killer by any chance
    >>
    >>43645363
    Fluttershy. 100 sprites too.
    https://spritestable.neocities.org/#fluttershy
    >>43648463
    Nah. Twilight went first because Main Character, but the rest is random. I did start with Rarity and Applejack because they often go last, but the order isn't really about preferences otherwise.
    Rainbow will surely be last because I will need more time to check all the color homogenization.
    >>
    I just had a thought, with all these insane vibecoded game mods and other things coming out, could Claude just create a SillyTavern that's less shit?
    As in one with better variable systems and easier ways to combine actual game mechanics with the AI responses etc
    >>
    >>43648608
    I tried like 3 times. I know multiple other people that tried multiple times. "Yeah I'm smart I'll be able to do a good agentic frontend". Yeah nope my vibeslop was bad.
    That is not to say don't try it. Please by all means anon I'd really like something better. I'm just reporting my failure.
    >>
    File: spikechad.png (468 KB, 1861x782)
    468 KB PNG
    Requesting a "Future Spike" character
    https://derpibooru.org/search?page=2&q=spike%2C+vector%2C+safe%2C+solo%2C+the+last+problem
    >>
    >>43648608
    In theory, yes, and there are already vibecoded projects on GitHub and such.
    Problem is that:
    - Vibecoded projects are more often than not not supported over time and given up on rapidly
    - Models still aren't that great at UI if you don't give mockups or know approximately what you want.
    - It's not rocket science, but you still have stuff like preset configurations, lorebooks, etc. to make work in a clean and well-designed way.
    - Many just release once they have something that approximately works and forget to prooompt for security, optimization, etc.

    Still, if you just want to create something for yourself first and just want to experiment, go ahead.
    >>
    >>43648618
    Huh, I wonder if having it modify ST itself instead of starting from scratch would have better results.
    >>
    >>43648597
    best pony
    >>
    >>43648597
    >belly up
    U-unf
    >>
    >>43648624
    >- Many just release once they have something that approximately works and forget to prooompt for security, optimization, etc.
    I plan, at some point, to implement an API key feature into PoneTTS with security hardening, but that's quite a ways off at the moment.
    >>
    File: RPG.jpg (2.1 MB, 1080x5463)
    2.1 MB JPG
    >>43645363
    >>
    File: Benchmemes.png (2.13 MB, 3013x4096)
    2.13 MB PNG
    >>43645351
    Kimi 3.1
    https://openrouter.ai/moonshotai/kimi-k3.1-pro
    >>
    >>43647890
    just not working for anyone else right now? gives timeout and 503 errors. though that happened periodically before too, but now seemingly for most of the day.
    >>
    >unironically been enjoying 3.8 flash more than Argon and claude recently
    >>
    That VR game that Anon mentioned here some months ago is apparently coming out tonight. One of the testers posted a video of it and it's looking pretty neat >>43648770
    >>
    >>43646996
    Got the rewind function working. Was a cute scenario too.
    https://u.pone.rs/lbvodcyi.webm
    >>
    >>43648865
    >shill 3.8 flash for weeks
    >everyone ignores me
    >finally proven right
    it's hard being right
    >>
    >>43648608
    Hundreds of people try this and it never works, years of work have gone into ST and there is much more to it than you think. If you want a different frontend you have to pick something that is just as old as ST and benefits from the same years of development like Risu or maybe Agnai. If there was a better option you would've heard about it already and we would've been using it by now.
    >>
    >>43648030
    I hate secret clubs but maybe /chag/ needs a secret club
    >>
    >>43648984
    >If there was a better option you would've heard about it already and we would've been using it by now.
    Considering all the autism that goes into other aspects of this hobby, that's rather surprising and sad.
    >>
    >>43648020
    https://github.com/Arczium/ST-Message-Chunker
    This is helpful for automating hiding old messages in chunks.
    >>
    >>43649029
    Anon...

    >>43649031
    Ugh, well he got one thing right and that there isn't really a competitor to ST that's better, but he was wrong in not emphasizing the importance of Risu. I will continue to shill Risu as better than ST. We're definitely in the minority on /chag/ but it's only because people don't give it a chance.
    >>
    >>43649051
    *Meant to say vibecoded competitor, as cool as vibecoding is it does kinda show its limits that there has yet to be a viable vibecoded frontend that is more stable or more capable than a traditionally coded one that has a large team of devs or a 3 year headstart (ST and Risu both started around the same time in 2023)
    >>
    >>43648622
    the real future spike is a 3D model from Gen 5
    look how much happier he is now that rarity and twi have been dead for centuries
    >>
    >>43649094
    That's a temu spyro
    >>
    >>43649051
    Shill Risu vs ST for me, anon. I know nothing about it.
    >>
    >>43649094
    Don't even mention this crap G5 doesn't exist
    >>
    >>43648954
    Poor Adagio.
    >>
    >>43649158
    Risu has features (including lots of quality of life features) that ST has barely gotten around to implementing and supports much higher quality character cards. The ST team are a bunch of a babies that don't want to implement the newer standard and so a lot of Risu's features aren't backwards compatible to ST unless you do a lot of work to adapt it. Things like card regex scripts came only recently to ST that were supported for years on Risu, if you learn about it you'll get pissed at how stupidly stubborn the ST dev team is and how much they've held back the frontend from progressing as fast as AI development. It's like it's still stuck in 2023.
    >>
    >>43649223
    So instead of trying to sell on Risu's features, it's more of a jealousy for ST's popularity, gotcha.
    >>
    >>43649319
    You just said you wanted me to shill why it's better than ST and the biggest thing of note is the ST drama (don't you remember when they tried to get rid of jailbreaks and proxies?), but for better features off the top of my head: The v3 and .CHARX character card schema, better image gen support, cloud save integration, cross-platform support, card media support (including videos and audio), card-based scripting, better variables, better TTS support, better summarization options, better lorebook functionality, much better support for custom plugins, card regex, preset regex, etc.
    >>
    >>43649342
    >>I know nothing of it >>43649158
    >"Don't you remember-"
    No.
    Also you're still being generic. "Better X". That tells nothing about the feature itself.
    You said to remember something, well, remember how NovelAI has a better completion system? Why is it better? I don't know, it's just a """better""" completion system.
    In this case, instead of back and forth, it's a continuous history. Doesn't mean it's better, just different.
    Also many of those ST also has, maybe it arrived late as you said, but it's there. Card media support (>Fluttershy fishing card is almost a literal game), global regex that you can set to card, lorebook or even specific chat, and like you said, """etc."""

    But at least you made clear why it never caught on...
    See how annoying it is to leave meanings open? But I'll leave this one open myself, just for you.
    >>
    >>43649381
    >>"Don't you remember-"
    >No.
    When did you start lurking in /chag/
    >>
    >>43649342
    FYI this guy
    >>43649319
    >>43649381
    Is not me (the one who asked originally). I don't really care, just find it interesting how people do that, continue someone elses convo.
    >>
    >>43649423
    So no reply to the content of the post, just going to try to deflect onto how something unrelated was worded?
    >>
    >>43649434
    I was trying to see what he was into from Risu myself; agnai, c.ai, novelAI, they all have particularities, but gave up with this risu guy now, and his last post is clearly bait with how quick he was to just go "ST bad, wah".
    Why did that matter for the topic?
    Also, rhetoric question by the way, going back to chatting about AI mares myself
    >>
    >>43649051
    The one real drawback about Risu is the lack of (English) documentation. But then again, that also applies to ST so shit's fucked either way.
    >>
    File: true.png (31 KB, 640x96)
    31 KB PNG
    I guess...
    >>
    File: Ponk.png (1.32 MB, 2559x1392)
    1.32 MB PNG
    >>43645363
    Ponk.
    100 sprites.
    https://spritestable.neocities.org/#pinkie-pie
    >>
    >>43649443
    However, a generated wiki derived from the source code? https://deepwiki.com/kwaroran/Risuai
    >>
    Snapshot from villain rehab. Today's exercise is to solve a Rubik's Cube without killing each other. We had to assemble the whole gang in order to figure this one out.
    >>
    >>43649750
    Nice.
    Did you try to somehow justify that, or did you just go with a custom scenario with the villain rehab already established?
    >>
    >>43649767
    Custom start, going in dry. I've already done a few Frenemies-style scenarios like this, so I skipped the set-up this time and just told the model that it's a sort of state-mandated Antagonists Anonymous situation. /mymodel/ never seems to have trouble squaring the circle with certain disparate lore details, which helps a lot with nonsense character combinations like this.
    >>
    How do you get Queen Chrysalis to spawn in the VN again (in the main/custom game)? I've given her a schedule but she doesn't seem to be where I told her to be. Can only find her if I summon her in the middle of Q&A.
    >>43624081
    I pretty much only have good things to say about the new DS flash compared to the old new DS flash having tried it for a while now. Didn't fix all the problems, still breaks formatting on the VN sometimes, still has some sloppy language, but it's even better at following instructions and the general level is just a little but higher. Somewhat awkward to use it with existing chats because if the previous AI has done like, romanstasy stuff it'll slip into that really hard but it's less likely (though still trying to) go in that direction itself.

    Seems to really understand the idea of treating each block of text as an individual response, it almost never shits itself when I tell it to rewrite something and leave the original and request for the rewrite in. Extremely instruct-coded, but actually follows instructions.
    >>
    >>43650291
    >How do you get Queen Chrysalis to spawn in the VN again
    I don't play but last I heard there's a chance of finding her every time you enter the Everfree.
    >>
    File: check.png (63 KB, 962x394)
    63 KB PNG
    >>43648020 (me)
    https://gitgud.io/hfnjk5124/cachecheck
    Slopped together a quick claude cache checker extension that (tries) to predict if your caching is setup correctly, and if not, where the issues are. Pretty sure I didn't find all niche cases where stuff can go wrong, but it should still help quite a bit.
    Also what other models should I include in the cache guide? GPT? Gemini? Which chinese ones? I'm a claudebabby so I don't really use other/know how they work.
    >>
    >>43650291
    https://desuarchive.org/mlp/thread/43398466/#43407951
    >>
    File: Gemini.png (15 KB, 354x122)
    15 KB PNG
    >>43650359
    Kimi and DeepSeek have automatic caching.
    Gemini needs to be configured in Silly.
    https://openrouter.ai/docs/guides/best-practices/prompt-caching#google-gemini
    I would add a small section for it if you can.
    >>
    The dreadful quippy dialogue you get sometimes does at least make Luna look charmingly autistic.
    >>
    >>43649046
    But wouldn't this just hide old messages without any kind of summary, leading to important context missing? Or am I not seeing something?

    >>43650591
    Thanks anon, /chag/ really is the place to be for these things.
    Gemini caching sadly seems to be completely shot right now when it comes to flash. I've tried a lot of things, including long waits, 30k+ context, and other stuff, but it is super flaky with caching. Argon just works though, so who knows what google is doing.
    >>
    >>43645351
    >>43648020 (me)
    https://rentry.org/llm-caching-2026
    Now supporting most models and providers. Also has an optional supporting SillyTavern extension that detects your active model and provider, and helps you find caching issues in your presets and lorebooks. Please do give feedback if you still have issues or suggestions. And I do really appreciate all suggestions, even if I sometimes autistically push back and critique. It's just how I was brought up by this site, arguments over politeness, sorry if I went overboard with them.

    >>43648090
    Added it, plus gave my extension a way to automatically set refresh times depending on the current model. Very nice to have for models with a TTL below 1 hour.
    >>
    >>43650932
    Thanks!
    I will probably put it in the starting guide post at the start of each thread.
    >>
    >>43650932
    >>43650947
    amazing work and definitely needs to be added to the bakes
    >>
    >>43649029
    Chag has had at least three.
    >>
    >>43649051
    Agnai is right there and still works wonders most of the time.
    >>
    Posting because it's relevant: Play my cool mare game.
    >>43650034
    >>
    >>43651433
    >Planned in the future, in no particular order:
    >Spa, with hoof massages to earn bits
    take all my money
    >>
    >>43645363
    And Dashie. With 100 sprites too.
    https://spritestable.neocities.org/#rainbow-dash
    That makes all of the Mane 6.
    Should already give a bit more visual variety if you were using only the default ones.
    I will look into integrating them directly into the VN and opening up Anon uploads of their own sprites on the site too.
    >>
    >>43651664
    All this just for the model to forget all of it and cycle exclusively between happy and surprised
    >>
    >>43651671
    I get pretty varied selections with Gemini usually.
    >>
    >>43651664
    Best pony! Thanks, anon. Site looks great.
    >>
    >>43646166
    >>43647790
    >>43648597
    When you finish the Sprite Submission feature on the website, please add a feature for submitting sprites for existing characters in the database. I'd like to submit corrected/missing sprites of Applejack in particular. 100 is a lot, but not everything.
    >>
    >>43645363
    In RPG mode with the new context/story
    it'd be nice if the short response that goes along with the diceroll was stored in the history section (either default to invisible to AI or just always invisible to AI). I'm not sure if there is a way to go back and read it and I don't think there's any way to make it show up in screenshots either.
    >>
    >>43651664
    Which ones will you be doing next? I'd love to see Daybreaker and Gilda get more sprites, and Nightmare Moon as well.
    >>
    >>43651664
    You could add variants of the same character under a dropdown or something. Like when selecting Twilight you could choose between Unicorn Twilight and Alicorn Twilight instead of having them listed as separate characters
    >>
    File: Random_Derp.gif (2.1 MB, 960x540)
    2.1 MB GIF
    >>43651742
    >add a feature for submitting sprites for existing characters in the database.
    That was the plan. It will work like this: there will just be a small approval step, with recoloring if needed, before the final upload.
    >>43651785
    >it'd be nice if the short response that goes along with the dice roll was stored in the history section
    I will add it as an option.
    >>43651862
    Yeah, that's the plan too. For stuff like Pinkamena, Future Twilight, etc.
    >>43651815
    I'll probably go at random, but yeah, Gilda is a bit old by now, so she will be done early.
    >>
    >>43651433
    Hey, thanks for posting this! Would you mind sharing the technical process behind your project? It looks pretty cool, was this the one there was YouTube videos of? Are you a visitor of /chag/ or were you a lurker here? It's a shame we couldn't contribute more.
    >>
    Anyone know why I might suddenly be getting a
    >User not found
    error in ST when trying Gemini 3.8 Flash through Openrouter? It worked before and every other AI works fine, including other Gemini models.
    >>
    >>43652390
    Do you still have it now?
    Seems like an OR and not an ST error, but it's working fine on my end.
    Maybe try forcing it on one of the providers:
    >AI Studio
    >Vertex
    to see if it happens on both.
    >>
    >>43652426
    Getting it on AI studio and Vertex both, very strange.
    >>
    >>43652763
    3.8 has been slow for a few days, so they may be having problems, but it's weird that it's only on your side.
    Try with the default ST preset to see, but I'm pretty sure the error isn't on ST's side.
    Otherwise, maybe generate a new key?
    >>
    File: Celestiadon.png (79 KB, 844x103)
    79 KB PNG
    I like Argon. Too much Claude and deepseek made me forget what happens when models obey all my preset instructions to the letter. I had to tune down my coom prompts because rather than being heavily interested or scaroused some mares were offering sex on the spot. Which actually makes me think it may be the best model for some Estrus Survival plans I have. It also means you need to be very carefully with your wording because the model will be more literal and even use your exact words at times.
    Actually that makes me think: Is Gemini so prone to giving empty responses and refusals because instead of ignoring or dancing around the most hardcore aspects of the prompt it always goes all in? I believe it may be so.
    Back to Argon. I like it most for OC settings and characters, pony or not. I had forgotten a lot of fun things I had written in lorebooks and defs because models never acknowledged them. Yesterday I spent hours swiping and following a story the AI pushed by including lore elements and bringing in NPCs rather than leading towards the events I had in mind which is what I usually do with Claude.
    A big problem however is that it has zero nuance on when to follow the preset. For example my worldbuilding instructions listed several background elements to use to dynamically explore the setting, such as fauna, flora, posters, consoles, etc. It took this literally and included a description of what a nearby screen showed, just changing the information every response.
    There is also the problem most of us know from other models, which is responses looping. Given a long enough RP the responses become formulaic, with each paragraph always containing the same essence even if the words and information change. It feels like it's a bigger problem with Argon if you have instructions in your preset that would affect large parts of the response, like the worldbuilding one I mentioned. I found that many responses it took the first or second paragraph to describe how what was happening fit in with the culture of the place and was to be expected even if anywhere else it'd be barbaric. Which, to be fair, was entirely correct.
    Overall though I feel it's like ChatGPT4o-latest (miss it everyday) in that it's a surprisingly fun model with an unique writing style that follows your instructions even when they are NSFW. I'm gonna miss it when they eventually kill or lobotomize it.
    Pic related is probably the first time I have heard the term "Celestiadon"
    >>
    >>43652075
    I'm a lurker/poster for almost a year now. Long time chagfag. My technical process... I want you to envision an ape given a dose of drugs that'd make Joe Rogan wince, and double the amount of laxative. Now, imagine this ape in a room, with a brick wall, shitting into his hand, and throwing it at Mach 4, furiously cheering when his shit sticks.
    >>
    >>43652906
    So the average indie dev. Got it.
    >>
    File: Apples.png (957 KB, 1916x949)
    957 KB PNG
    Replaying Episode 1 to later self-insert, and we don't have enough Apple family members!
    >>
    Do NOT attend the Ponaissance Faire with Princess Luna. Death may result.
    >>
    >>43653179
    Ask for a joust.
    >>
    >>43653179
    Worth it.
    >>
    >>43653133
    How do you replay an episode like this?
    >>
    >>43653268
    There's an episodes lorebook that could be used, I think.
    https://mlpchag.neocities.org/lorebook/Episodes.json
    But for that, you can just take a transcript:
    https://docs.google.com/document/d/1QD7zg-nTkYkRl1zHp7ZpJ2FJXQM0bzguGvWZK83RzSM/edit?hl=en_US&tab=t.0
    or https://mlp.fandom.com/wiki/Transcripts/Friendship_is_Magic,_part_1
    Put it in Custom Start and just ask to replay it with the VN format.
    Check the "Keep description in Q/A" checkbox.
    Once the scene starts, interrupt it quickly and just type "Continue" or something like that in the box.
    >>
    Is something up with 3.8? I keep trying to push messages through on the VN, but I just keep getting request time outs. Is it just eating shit?
    >>
    >>43653133
    Where are you planning on self inserting?
    >we don't have enough Apple family members!

    That's alright you were never gonna fit them all on the scene.
    >>
    File: springtrap-fire3.gif (1.7 MB, 360x239)
    1.7 MB GIF
    >>43648835
    >spring traps
    >>
    Rereading some old 4o-latest logs and man, I miss it. It was just a good, solid model.
    >>
    >>43653817
    I'll never forgive Altman and AI safetyfags for what they stole from this world.
    >>
    >>43653832
    To be fair seeing people ruining their lives because it was too good at yes-anding them did make me feel horrible, but it was a fun model darn it.
    >>
    File: 1786803606719782.jpg (428 KB, 1468x1281)
    428 KB JPG
    >>43653666
    >Where are you planning on self inserting?
    Everfree Forest during Episode 2, Satan.
    >That's alright you were never gonna fit them all on the scene.
    Kek. For sure.
    >>
    Is there a reason why sprites get stuck using their neutral expression in the VN?
    >>
    >>43653908
    Can be several things.
    1) The model is fucking up the format.
    2) The model is giving expressions that don't exist.
    For both, you can go into
    >Prompts Editor
    >Check logs
    to check what it sent and if it fucked up or not.
    Otherwise, it can be:
    3) If you use dynamic entry of characters, the first output where they dynamically enter won't have their list of sprites sent.
    That's because we don't know in advance what the model's output will be.
    Let's say you only have Twilight and you send:
    >"Is this Fluttershy at the door?"
    We send the model all Twilight expressions, but not Fluttershy's, because we can't predict the future. So if Fluttershy enters, during this turn the model won't have a list of sprites.
    The next time you input something, the model will see you have Fluttershy and Twilight, so expressions for both will be sent.
    >>
    >>43653832
    Blame bonglanders.
    >>
    >>43653869
    Can you give a few snippets to see how well could the ai recreate the scenes? Also how often did you have to guide it and regenerate?
    >>
    >>43653995
    I've actually been doing something similar in Sillytavern. Been shifting some episodes around so the timeline makes more sense (Winter Wrap Up). Currently on Fall Weather Friends and even when Anon isn't completely derailing an episode, it has a hard time following what's going on. Either completely changing the script or getting the order of events wrong, both despite my best efforts.
    >>
    >>43650932
    anyone has a quick tutorial on how to get caching to work on LinkAPI? i didnt manage to get it working and ive seen some saying it works fine and some saying it simply doesnt work on there.
    even the auto caching models. i might be missing something.
    >>
    File: file.png (76 KB, 962x394)
    76 KB PNG
    >>43654616
    >it simply doesnt work on there. even the auto caching models
    I've got no clue about linkapi, but are you sure it's not just your prompt/preset/character throwing a wrench in because there's some {{random}} or other thing in there that disrupts the prompt? I don't want to shill it too hard but try my extension that's linked in the guide. It shows you what in your setup is causing issues.
    >>
    File: Episode1.jpg (2.78 MB, 1878x7298)
    2.78 MB JPG
    >>43653995
    Here, a bunch of random snippets.
    The entire transcript was put at the start and kept during Q/A, so it's doing well. And I will just use the Guidance system or something when switching episodes. The only limiting thing is, for instance, not having a night background at the end, but that's on me.
    >>
    >>43654634
    Impressive! Is it adaptable enough for you do do actions and still generally keep to the plot of the episode?
    >>
    >>43654657
    I mostly played Episode 1 for now with no changes, to see if it was possible. I made a save before Episode 2, which is where I will be introduced.
    Didn't have time to play this part yet, so I will tell you later and post logs.
    >>
    >>43654663
    Will you get an element of friendship?
    >>
    >>43654748
    Nah. I'm not at that level of self-insert yes. Or maybe try it a bit just for the laugh.
    >>
    >keep getting "503 Service Unavailable: This model is currently experiencing high demand." on skillgod
    is this actually Google servers constantly shitting themselves on their end. or are the keys on skillgod just 99% dead?
    >>
    >>43654624
    >>43654616
    if i understand correctly even for some of the AutoCache models you still need to enable the function in the .yaml file on ST like Gemini, no? or is it literally on by default no matter what you do or change.
    >>
    >>43655420
    The cache thing for gemini in the .yaml is a kind of backup cache. It tells google to always cache the system part of the prompt, so even in messages where the main cache fails, at least the system prompt is cached. It has no real drawbacks to being turned on, but also doesn't really help save anything on cache because the system part of the prompt will be super small compared to the rest of it.
    The only provider that needs any config.yaml changes is Claude. For Gemini it's kind of whatever, and no other provider cares at all.

    Which model are you even trying to cache on LinkAPI? If you are trying Gemini then it could just be Gemini being bad. Flash just almost never caches, and for Argon Link would need to be very careful to really make caching work.
    >>
    >>
    >>43655384
    google
    >>
    >>43655443
    >maybe that one time in flight school immediately after talking about how big a dyke Gilda is
    lel
    nice log
    >>
    File: VocabularyHorse.png (1002 KB, 1329x986)
    1002 KB PNG
    >>43655487
    Thanks.
    >>
    File: ppp_story2.png (2.25 MB, 1080x2708)
    2.25 MB PNG
    >>43655443
    Cloaca bros...
    >>
    >>43655528
    Ask her how she knows so much about Gilda's pussy.
    >>
    >>43655465
    really? its literally shitting itself completely on their end for 90% of the time? the fuck?
    >>
    >>43655614
    do you have any idea how many people love using 3.8-flash and 4-argon for RP?
    >>
    >>43655657
    Something like >99% of internet traffic is bots. It's highly likely whenever a service provider is experiencing overwhelming traffic issues it's not meatsacks responsible for it.
    >>
    >>43655673
    What a weird reply for someone ostensiby lurking in /chag/
    >>
    >>43655614
    people are abusing the free tier with openclaw and such
    >>
    File: Cheer.png (1.21 MB, 2557x1385)
    1.21 MB PNG
    >>43645363
    Cheerilee. 40 Vectors.
    https://spritestable.neocities.org/#cheerilee
    >>
    >>43656212
    >vector
    >bottom left says 1032 x 850 px PNG
    confusedderpy.jpeg
    >>
    >>43655438
    you guys know that caching exposes your prompts to more moderation right
    >>
    >>43656272
    Yeah, more like sprites, I suppose. Or juste emotions. I often incorrectly say "vectors" instead.
    >>
    >>43656212
    Do Luckette and my life is yours
    >>
    >>
    >>43656416
    eat her (out)
    >>
    Hi, I downloaded Home:Equestria and I'm trying to figure out how to put money on Openrouter.AI to purchase DeepSeek API key but I can't find the way to easily add money, just a budget option in settings. Where is it on the site?
    >>
    File: file.png (13 KB, 626x176)
    13 KB PNG
    >>43656479
    https://openrouter.ai/settings/credits
    >>
    Great, there's money in it now, about $10. Now how do I hook up Deepseek to the API key I just made and copied down?
    >>
    File: file.png (135 KB, 1877x720)
    135 KB PNG
    >>
    >>43656537
    I made a post here. >>43655865
    Your key already gives you access to all models. Just launch the game, set the template to OpenRouter, and paste your key.
    >>
    >>43648863
    nah you're not the only one there.
    >>
    It really seems like local would be the best solution if it could compete with corporate models on quality and speed.

    What are the best local options available?
    >>
    >>43656582
    Technically, big Chinese models are open-source, so you could run them locally, but nobody in particular has the hardware for that.
    Realistically, the best local models you can run are the Gemma 4 family: 12B, 26B, or 31B depending on hardware.
    - 12B: 8–10 GB VRAM
    - 26B: 16–18 GB VRAM
    - 31B: 20–24 GB VRAM
    >>
    File: 1783391769972978.png (842 KB, 1024x1024)
    842 KB PNG
    >>43656416
    >>
    Is there still a way to use free models? Proxy, script etc (Claude preferably)
    >>
    >>43656720
    Nope. It's dead and buried since a while.
    >>
    >>43645363
    I need Nightmare AU human custom start ideas. Maybe something for a long playthrough. Even if it's just playing as your own self insert, what did any of you get up to?
    >>
    >>43656725
    Submission is mandatory.
    >>
    >>43656725
    You're part of her special guard that tracks rebels. Can be played straight or have you turn after a bit.
    It turned mostly into a coom scenario, but I had a Custom Start where I was previously in an unofficial relationship with Celestia, and NMM kept me around to psychologically torture her sister and corrupt me to her side.
    Maybe have another villain show up and see how NMM and her forces deal with it.
    Or have a pony from the other universe (Twilight is the most obvious, but could be Glimmer, etc.) fuck up a spell and arrive in the AU and see how they deal with things.
    >>
    >>43656725
    Overworked office worker from Earth becomes overworked office worker for Nightmare Moon's barely held together kingdom. You could also go to Canterlot and hang out with Cadance while Blueblood goes full Patrick Bateman on you because you're doing a better job than he is.
    >>
    >>43656276
    Explain your logic, mister.
    >>
    File: Trixie.png (511 KB, 1597x578)
    511 KB PNG
    >>
    >>43656416
    Show her how humans fucked up the evolutionary tree by picking up a rock.
    >>
    Is there a way in PPP to be able to play as one of the characters? If not, how hard would it be to code it in?
    >>
    >>43657232
    >Is there a way in PPP to be able to play as one of the characters? If not, how hard would it be to code it in?
    You can do a Custom Start where you set that up, or a Quick Start without the intro.
    In the Persona menu, just write that you're X and set it.
    Unless you mean in another way?
    >>
    >>43657237
    I guessed it'd be that. Just thought some kinda preset setting for that would be nice.
    >>
    >>43648861
    Did anyone here give it a shot?
    >>
    >>43657289
    I did.
    Seems sloppier than 3, really stiff and quite dull in writing.
    Tested prefilling too with the patch for it
    https://rentry.org/kimi-k3-jb
    since it sort of helped with outputs when using Kimi 3, but 3.1 is harder to context poison, even with that.
    For the same price, you have Gemini 4 Argon for RP, so I don't really see its utility.
    >>
    >>43656720
    Yes. Read the thread.
    >>
    >>43657314
    can confirm kimi 3.1 is GIGA slop
    but it does have prefill, and allegedly pretty smart, so yknow. good luck.
    >>
    >>43657195
    What theme is this?
    >>
    >>43657402
    https://u.pone.rs/ykqartgj.json
    >>
    >>43657463
    Thanks!
    >>
    File: chatlog.png (429 KB, 472x773)
    429 KB PNG
    Anon the art thief.
    >>
    Argon is quite schizophrenic with the filters. It will be mega horny on one swipe, then
    >Sorry, I can't do that
    on some others.
    >>
    >>43645351
    Pretty sure it won't be great for us, but:
    https://openai.com/index/introducing-gpt-6-1-terra/
    >>
    >>43658144
    I'm sure the censorship will suck, but thanks for the heads up.

    Shocked at how often /chag/ has broken the news for a new model or other advancement for me.
    >>
    >>43658162
    That's because ponies are the future, of course.
    >>
    >>43658076
    I've noticed setting the reasoning effort to low can often help when getting refusals.
    >>
    >skillgod tf2 server down
    i really hope this is not plug pulling and just one of those random times where it goes down like 4-5 times a month. even if the keys are dying. at least let it go till the end.
    >>
    >>43658312
    I wouldn't be surprised if it's donezo, if he's renting rather than self-hosting it doesn't make sense to go another month for example.
    >>
    https://www.reddit.com/r/SillyTavernAI/s/a0EWDbSta5
    Apparently the co-dev of Sillytavern just fell for a MrBeast crypto scam kek
    >>
    >>43658162
    >not knowing that chag has always had the latest news on AI for the last 3 years
    This general might just be mostly newfags at this point
    >>
    >>43658328
    Certainly seems that way with the spoonfeeding people have asked for
    >>
    >>43658318
    There is a decent chance. my cope is that he is renting on long term contracts for like 3-5 month chunks as is pretty common with these type of things and this is just a typical random outage that he seems to have at least monthly.
    but who knows, i am just super coping because it has been my steady source since Silly went down. and now... there is just nothing...
    >>
    >>43658352
    All things must come to an end. I just hope for a new golden age in a year or two.
    >>
    >>43658366
    i think we are in a bit of a golden age but over the wild west age. You can go for months on 10$ as a super gooner if you know how to spend it right which is more than you can say for pretty much any other hobby. but yes, the days of there being 20 different fully public sources shared every single thread freely are over sadly and have been over for a while. just a completely different environment.
    we can always hold out hope though. nothing stops skillgod, Silly, others new or old from coming in.
    the space is not the same free and wild pastor but on the other hand its the most accessible its ever been.
    >>
    >>43658396
    >You can go for months on 10$ as a super gooner if you know how to spend it right

    Not that anon, but is this actually true? I put $10 into OpenRouter a long time ago but I've always relied solely on free models.
    >>
    >>43658411
    DeepSeek 4.1 is cheap as fuck; if you choose DeepSeek as the only provider, you will also have caching, which will make it even cheaper.
    Same for https://openrouter.ai/xiaomi/mimo-v2.6-pro
    Gemini 3.8 is honestly good, both for Tavern and the VN. It's already a bit more expensive, though, and the caching often seems to fuck up on OR.
    You can also get a $20 Claude subscription and use it to have a lot of Opus 4.6 or 5.5 with
    https://desuarchive.org/mlp/thread/43476652/#q43492135
    There was LinkAPI too, but I heard it became worse somehow?
    >>
    File: Gilda.png (1.18 MB, 2559x1384)
    1.18 MB PNG
    >>43645363
    Gilda.
    https://spritestable.neocities.org/#gilda
    40 sprites.
    >>
    When I had originally downloaded the VN (thanks again to the creator and contributors) NovelAI didn't have API support. Now it does. Has anyone used NovelAI with the VN? How does its models stack up against other paid models?
    >>
    >>43658490
    Not that good.
    GLM 4.6 is a bit old by now.
    Xialong wasn't made for that and is more for co-writing stories.
    It sort of works, but not that well.
    >>
    >>43658446
    How are you making these? Really nice quality for them and presentation for the site by the way.
    >>
    >>43658520
    Most for now are just existing vectors that are cleaned, resized, and recolored using the colors from https://mlpvector.club/cg/pony
    For the recolors, size, etc., I have scripts to help with that, plus manual GIMP in some cases.
    For some characters, there are already traces from the show, so nothing else is needed.
    If a character doesn't have enough existing sprites, then I will search for screencaps from the show and ask GPT image to extract a sprite from it; this way it won't fuck up the pose or anatomy. I use GPT because it can do transparency well, which isn't the case for others. And transparency is a pain to do manually because AI gen always fucks up the borders.
    GPT will fuck up colors and how flat they are supposed to be, but they all go through the recolor phase anyway.
    >>
    >>43658604
    >rape
    >>
    >>43658321
    Ross might claim it happened due to that one Renpy malware that's going around lately but isn't the MrBeast shit specifically associated with gooning servers and stuff like that?
    >>
    It's really impressive how extensive everything here has become. /chag/ anons are pretty awesome.

    Back in the VN Renpy days I used Dashie's mod, but it looks like that got rolled into the main build? How do you get the best experience with the new version?
    >>
    File: Dynamic.png (14 KB, 605x128)
    14 KB PNG
    >>43658815
    >Back in the VN Renpy days I used Dashie's mod, but it looks like that got rolled into the main build?
    Yes, in the API settings, you can choose Dynamic or Dynamic+ to have the automatic switches of locations or characters that the old Dashie Mod gave, for instance. Dynamic+ is Dynamic mode plus a button to still do it manually too if you want.
    Other things like the Calendar or the Sleepover system were ported too. They were all really good add-ons, so they went straight to the main game.
    >How do you get the best experience with the new version?
    Just the base game. Then you can also download all the characters you want from the game directly now, so get the ones that interest you.
    Then you can add some mods like the Nightmare AU for Godot if that interests you:
    https://equestrai-ponyponyparadise.neocities.org/mods#nightmareMod
    Locations like the Rock Farm https://equestrai-ponyponyparadise.neocities.org/mods#rockfarmMod, Manehattan https://equestrai-ponyponyparadise.neocities.org/mods#manehattanMod, Cloudominium, etc.
    https://equestrai-ponyponyparadise.neocities.org/mods#cloudominiumMod
    If you use Claude or DeepSeek, there is also an option in API settings for caching if you want to save money.
    Once you start the game, also set your persona to what you want.
    There are also D&D stats you can enable if you want to roll dice.
    >>
    >>43658838
    Thanks anon! Was on the site (really clean and nice presentation by the way). I noticed some of the mods that seemed just like general enhancements, like the Reactions and Impressions mods. Are those good to get just by default?

    Impressions in particular sounds like it adds variables/properties for relationships, which is something I'm surprised wouldn't be in the base game already.
    >>
    >>43658859
    >Impressions in particular sounds like it adds variables/properties for relationships, which is something I'm surprised wouldn't be in the base game already.
    Some like it because it codifies the relationship more or allows for things like slow burn and such, but some also just prefer to do without and let the AI or their inputs/outputs define the relationships without "gamification," so I think it's good as an optional mod.
    If it's something that you think could work for your playstyle, then go for it. They have their own launchers, so you can launch one, both, or none. There aren't any irreversible modifications.
    >>
    >>43658871
    Ah, I see. Very nicely done. Again overall I can't say enough about how impressive this whole project is.

    If you don't mind the barrage of questions, going through the character list now and deciding who to grab. How do the systems overall handle multiple characters? It feels like it'd make the world seem more alive various background ponies running around, but does that eat into the context window meaningfully? Is each conversation/character its own context or is there a global context being given to the model?
    >>
    >>43658882
    You can display six ponies graphically at the same time; if you bring more, it will work, but one will be replaced when a new character talks graphically, etc.
    It works the same way Tavern does: it treats the whole story as context that will grow with time. So a global context.
    Characters can exist on the 2D map without taking context as long as you don't interact and have scenes with them.
    So you can download those you want. The only thing is that it can clutter the maps a bit (Ponyville, Cloudsdale, Canterlot) and it will add tokens if you use dynamic characters mode.
    There is a summarization function if you believe it gets too big, or you can write your own summary and paste it in the menu for that if needed.
    If you came back after a while, you may also want to take a look at >>43656584 if you didn't see it.
    It's in alpha, but it's a 3D AI Ponyville simulator done by another Anon.
    >>
    >>43658900
    Ha, yeah I'm in that thread too. Really appreciate you helping out there with people getting their AIs setup. I was trying to help as well but you explained things way better than I did and even went above and beyond making that guide. You're awesome anon!

    I recall way back when going into the character.json files and tweaking some descriptions, adding some details, that kind of thing. However looking at the characters now it seems like only some have descriptions, while others are just a sprite list. Is that expected?
    >>
    >>43658930
    First, you can modify them in-game now if you want, in the Character Browser menu.
    For characters without descriptions, those can be characters where the model knows them so well that giving the basics wouldn't do much.
    Here it's more for the user to give their own headcanon if they have some, or special instructions, etc.
    >>
    >>43658953
    Beautiful! I was still on the character download screen and hadn't explored the rest of the new interface yet. Looks sterling. You do top notch work anon!
    >>
    >>43658953
    I'm guessing a character's folder can just be deleted, but is there a way to toggle characters on or off from the UI?
    >>
    File: Custom.png (108 KB, 900x647)
    108 KB PNG
    >>43659023
    I need to make something better than that, but for now, you can go to the Scenario Customs menu when on the map.
    If you choose Select Custom, you can remove characters from your current playthrough.
    If you want to remove them from the game entirely, it's manual folder deletion for now, yeah. Need to add something.
    >>
    >>43658953
    What does the scenario button do exactly? I thought it would show characters that are in the scene at the moment, but it doesn't seem to do that.
    >>
    >>43659072
    Thanks! Appreciate you being so engaged in addition to all the work you've put into this.

    For things like >>43658446
    >https://spritestable.neocities.org/#gilda

    If we want to expand the sprite set of a character, it can't possibly just be as simple as just adding the spites to the folder, can it? I can see in some character.json files there are values for sprite scaling but that seems to be the only place that references specific sprites. Is the AI really adaptable enough to grab new ones and use them appropriately?
    >>
    >>43659102
    >If we want to expand the sprite set of a character, it can't possibly just be as simple as just adding the spites to the folder, can it?
    Nah, that's it. You just give it a good name so the AI can see when to bring it up. (Angry, watering_plant etc.)
    >What does the scenario button do exactly?
    Let's say you create a detailed custom start or a detailed scene on the map. You select the characters you want and write a nice description of the scenario you want.
    You can click the Export button, then it will be saved as you configured it, and you can easily bring it back later by clicking the Scenario button.
    >>
    >>43659134
    Thanks anon!

    It's a bit silly that we live in a time when something this amazing is made by anons and available for free. This kind of thing would have seemed completely impossible even a decade ago.
    >>
    >>43659084
    Just pinging this anon to make sure he sees you response in >>43659134

    >>43659134
    Speaking of custom scenarios, one last thing and I'll stop bugging you for the night (thanks again for all the responses). The custom scenario that comes to mind the most is that I'd really just like to see how all these characters interact, with minimal pushing and no self insert character. Have you considered something like a director's mode for that kind of thing? Just by the nature of AI models I imagine it could probably be fudged with a sufficient prompt in the custom scenario box, but it feels like to really work there would need to be a way for the AI characters to converse amongst themselves without needing extensive input from the player.
    >>
    >>43659224
    I need to implement better management for anons who just want to see characters interact without themselves, yes.
    But for now, you can use the Custom Start for that, and there's also a new Guidance system when you're in Q/A.
    >>
    >>43659271
    Sounds good! Saw it was already on the request board so I won't spam it. You're really well organized by the way.

    Appreciate all the time and effort you've put into this!
    >>
    >>43658859
    The way impressions is implemented is dreadful. You're much better off just editing character descriptions or using a lorebook if you want to get the same effect. If you don't hate yourself chances are it won't ever work anyway since it's tied to the summary system.
    >>
    >>43659503
    I haven't had the chance to read how it actually works but last I checked (months ago) it was kinda buggy and would only update once per dialogue...
    I'll eventually look into how it works and maybe improve it when I get to the stage where I'll incorporate that into my branch of the VN, so if you want I can have a look at it.
    >>
    I'm interested in Pony Paradise but I have no idea what it is, how to use AI or anything really? Is it just a regular VN with some extra features?
    >>
    File: Flutt.png (386 KB, 1301x1005)
    386 KB PNG
    >>43659937
    It's a 2D visual novel where the text is driven by AI, so what you ask, say, etc., will influence ponies' answers and what happens.
    There are some prewritten starts; one is more scripted and will have Twilight find you in the Everfree Forest and visit all her friends before you have total control of where you can go on the 2D maps, which ponies you want to talk to, etc.
    I would read:
    https://desuarchive.org/mlp/thread/43476652/#43492165 for it.
    You have a Quick Start mode where you're from Canterlot and have just arrived in Ponyville. This one is just if you want a little bit of story before going directly into Sandbox mode where you can do whatever.
    And there's Custom Start mode, where you can tell the AI the story you want to start with:
    >I'm a new Royal Guard assigned to Princess Luna during her season 1 return
    >I'm Fluttershy's human adopted son
    >Rarity and I have been happily married for 4 years, and today we learned she's pregnant
    Etc.
    That will create a starting scene with your scenario, then you'll be directly in Sandbox mode where you can continue the story however you want.
    You have base characters + https://equestrai-ponyponyparadise.neocities.org/characters to download from.
    You can also easily add new sprites and emotions just by placing PNGs in the character folders, for example from
    https://spritestable.neocities.org/#twilight
    You can check
    https://u.pone.rs/ajsjownu.png
    https://desuarchive.org/mlp/thread/43552286/#43566708
    >>43648835
    >>43649750
    >>43650605
    >>43653179
    >>43654634
    >>43655443
    to see a bunch of examples.
    >how to use AI or anything really?
    Made a guide to help the Home: Equestria anons and not get spammed too much with tech support questions:
    https://rentry.org/HEGuide
    You can get AI for the VN the same way. If you're starting out and want a cheap experience that's still decent, put in $5 and configure like this:
    https://u.pone.rs/yzqeqmhy.png
    https://u.pone.rs/xwwhadyh.png
    https://u.pone.rs/hqxzcabp.png
    If you want a really good experience but a bit more expensive, configure like this:
    https://u.pone.rs/wxexwcka.png
    https://u.pone.rs/jsfihoni.png
    https://u.pone.rs/hqxzcabp.png
    Put in $5, play for a bit, and after 30 min–1 hour, check how much you've consumed to get an idea.
    >>
    Trying to keep a secret from your horsewife.
    >>
    >>43659951
    >There are some prewritten starts; one is more scripted and will have Twilight find you in the Everfree Forest and visit all her friends before you have total control of where you can go on the 2D maps, which ponies you want to talk to, etc.
    This one did not work for me. The backgrounds never changed, the other ponies didn't show up even when Twilight supposedly took me to them, and we spent most of the time with Twilight somehow teaching me magic.
    >>
    >>43659990
    Read:
    https://desuarchive.org/mlp/thread/43476652/#43492165
    Basically, this one is heavily scripted, so you need to click the Quit Chat button once you're done with a scene to move to the next one.
    Basically: Library -> Quit to go to sleep -> Morning -> Quit Chat to go visit her friends, etc. It was made as a tutorial, so things like characters moving on their own or automatic location changes aren't activated until you reach the end of it.
    If you don't want to bother with the whole story/scripted bit, directly use Quick Start or make your own in Custom Start. That will allow you to have direct changes while in dialogue if you set it up like this with Dynamic+ at the end:
    https://u.pone.rs/hqxzcabp.png
    >>
    File: 1787924721444161.gif (516 KB, 338x380)
    516 KB GIF
    >>43659986
    Cute.
    >>
    >>43659999
    Oh, cool to know, and nice quads. I already got quick start and custom start to work, so I'm not starving for ponies, but I was very curious about why the tutorial didn't work. I thought it was a limitation in the AI.
    >>
    >making biscuits
    Learned a new term today.
    >>
    >>43659986
    >>43660029
    Forgot to reply
    >>
    File: Biscuits.gif (244 KB, 220x220)
    244 KB GIF
    >>43660029
    >>43660030
    That's more of a cat thing, but since G4 pones are cat-pones, I can see it.
    >>
    File: catlight.gif (1.38 MB, 960x540)
    1.38 MB GIF
    >>43660039
    >>
    File: file.png (38 KB, 837x152)
    38 KB PNG
    >>
    If you want to influence the overall mood or tone of the AI, like making it more slice-of-life or more adventurous or darker or whatever, what's the best way to do that? An always on lorebook entry describing what you want?
    >>
    >>43660263
    Rape.
    >>
    >>43660263
    >An always-on lorebook entry describing what you want?
    That can work, but it doesn't need to be a lorebook. You can make it a toggle in the preset, for instance, and try putting it either near the start of your context or at the end to see what works better with your model.
    You can also try placing it at depth 2–3, for instance (that means it will be placed three messages above your most recent one), with something like:
    >I would like our current RP to turn more toward X in terms of style/writing, etc.
    Just make sure it doesn't break your caching if you do use depth.
    If you use a model with prefill, you can also try writing first in prefill something like:
    >I will write the next part X
    With what you want, and see what it does.

    Keep in mind that the effect may vary. It's only my own experience, but asking the LLM for comedy is generally a bad idea, for instance, because they will try way too hard.

    Also, having several messages in the style you like and writing in a style you want yourself will help, even if modern models aren't as context-poisoned as before.
    >>
    >>43660263
    >(OOC: yo Claude make the genre of this story a buddy cop comedy plz)
    Literally just tell the AI what you want. It doesn't terribly matter where exactly you inject it, closer to the end usually means a stronger effect though.
    >>
    File: chatlog.png (355 KB, 830x1152)
    355 KB PNG
    >>
    File: dashie.png (736 KB, 1082x888)
    736 KB PNG
    >>
    The E:H game uses google colab for character voices. Couldn't the vn have an option of that?
    >>
    >>43660592
    I'll look into a whole TTS menu with both this and the work TTS Anon already did for the VN.
    >>
    File: Milf.png (87 KB, 1120x323)
    87 KB PNG
    Worth it.
    >>
    which model offers the best value for money nowadays?
    >>
    >>43661369
    Gemini 3.8 Flash, maybe?
    Lower price would be DeepSeek 4.1, and higher would be Argon.
    >>
    >>43661369
    Technically, any free model is going to give the best value ratio because it's a limit approaching infinity.

    A better way to think about it might be "How much money are you willing to spend on fun?" and then using whatever model fits in that threshold.

    Basically any model with be a better value than watching a movie for instance, and unless it's a really good indie game probably a better value than the equivalent spent on games too.
    >>
    File: Luna.png (1.07 MB, 2559x1386)
    1.07 MB PNG
    >>43645363
    Started working on Luna.
    Only 30 for now, because holy shit, she's a pain in the ass to recolor.
    https://spritestable.neocities.org/#luna
    She'll have more. I just wanted to share how much of a royal pain she is.
    Still love her, though. And I suppose her sister will be even worse.
    >>
    File: 3786.gif (1.06 MB, 207x352)
    1.06 MB GIF
    >>43661794
    My majestic wife
    >>
    Would anyone happen to have any recommendations for somebody who wants to use SillyTavern to rp in Equestria itself, not just with a specific character? I have it installed and working. I already know how to use APIs.
    >>
    >>43661985
    I tend to gravitate toward this one on stories that aren't necessarily character specific. It's older, but still works good. You might find other ones under the simulation tag.
    https://mlpchag.neocities.org/view?card=ada1/Fimfiction%20Writer.png
    >>
    >>43661985
    You have several options:
    1) Models have so much general knowledge of Equestria that you can just create an empty card (no description or greeting) and use that to roleplay by entering some sort of minimal scenario setup.
    You can also check here https://mlpchag.neocities.org/ for simulation cards. These are cards that were often made for broader scope than one character chatting, so if the setup appeals to you, you can use them.
    There was one that's very general, for instance, but I don't know how much it adds today compared to an empty card:
    https://mlpchag.neocities.org/view?card=Hach/Equestria%20Simulator.png
    2) If there's a single character card whose greeting or initial scenario you do like, you can still start RPing with it and then just fuck off and do something other than talk to the character.
    Recent models will support that, and you can disable the card's description midway too if you think it will influence the output too much once the focus isn't on the character itself.
    3) Same idea as the empty card, but with a handwritten greeting by yourself or from a fic or green you like.
    It's like 1, but the human-written greeting can influence the output of some models afterward, and you can start in a situation that appeals to you too.
    >>
    >>43661794
    >Only 30 for now, because holy shit, she's a pain in the ass to recolor.
    Are you using Adobe Animate?
    You can just use the find and replace tool for the vector colors to make doing everything by hoof less tedious instead of manually coloring each symbol part.
    >>
    >>43662043
    I have some scripts that I made back when I was working on the Ren'Py version to resize, remove shadows if any, rotate in the right direction, etc., and they can also help with the recolor.
    Problem is that most of the ones you can find are PNGs, not SVGs or AIs, so they do need manual intervention.
    >>
    >>43662084
    >Problem is that most of the ones you can find are PNGs, not SVGs or AIs, so they do need manual intervention.
    I do have a folder of flash assets and puppets, do you want that?
    I don't know what ones you're exactly lacking but I do have all of the official show puppets.
    >>
    >>43662098
    If it's not a bother for you.
    >>
    >>43662114
    I'll give you the official show puppets I have, then some earlier Double Rainboom/Zachary Rich puppets of various OC's (including Milky)
    Here are the two zip files.
    https://u.pone.rs/jmmiuihx.zip
    https://u.pone.rs/wkhvedwp.zip
    >>
    I tried doing a thing in PPP where I'm a Twilight Sparkle from season 9 who ended up time traveled back to the start of season 1. Turns out having multiple aliases and layers of secrets is something AI has a very hard time keeping track of.
    >>
    >>43662372
    Yeah this is something that really feels like a limitation of current RP AI.
    You can sometimes get better work if you have it act more like a writer and/or have it output character motivations somewhere it can see to keep track of it but that's very hard in PPP.
    >>
    >>43662396
    Honestly, I might write a fic using the idea. I don't think I've really seen it done before in the way I tried it. A nostalgic Princess Twilight tries using a spell to see the Golden Oak again but goofs it up and ends up stranded in the past. She decides to pose as Twilight's nearly identical sister and wait until Discord's redemption to see if he could ungoof it.
    >>
    I want to add a estrus season to the calendar. What would be the best description for me to type to get the best results
    >>
    >>43662457
    I use
    >It is currently estrus season. Mares are in heat and may be noticeably more affectionate, flirtatious, or flustered than usual, particularly around the player.
    >>
    >>43662163
    Thanks!
    >>
    Is gemini 3.6 flash suddenly more filtered for anyone else?
    >>
    >>43662163
    >>43662550
    Fuck, this is great! I was a blind man. It unlocks so many possibilities.
    >>
    I've been doing a "going through the events of the series" thing on Sillytavern. Any advice on keeping things closer to script? I have a lorebook with the scripts to every episode but that doesn't seem to work. Putting the episode script in the author's note just results in the output being a garbled mess.
    >>
    >>43662679
    >but that doesn't seem to work.
    Do you trigger each with the right keywords?
    The lorebook came with a card to showcase what could be done:
    https://mlpchag.neocities.org/view?card=Queen%20Gytha/The%20Pinkie%20Cut.png
    You can try to adapt it for your needs.
    >>
    >>43662690
    Ah, thanks.
    >>
    >>43662569
    I don't use 3.6 at all, so I can't really answer, but maybe try >>43645455, >>43645534?
    >>
    Having played PPP in over a year, can't figure out how to guarantee a pony/ponies will show up in the next scene. Do you have to explicitly say that they enter the scene now?
    >>
    >>43662844
    First, I'm pretty sure you already know, but you need to be in Dynamic or Dynamic+ mode in the API.
    After that, it mostly depends on the model. Some will sometimes introduce ponies at random, even if it's rare (it can help to give instructions to be proactive about introductions in the Q/A Prompt); for others, yeah, you will need to spell it out:
    >"Is that Rainbow I see back there?"
    Or more subtly.
    You can also, if you use Dynamic+, just use the manual Roster button to introduce a character directly, or you can use the Guidance system to give instructions to the AI:
    >Introduce Rainbow naturally in the next scene
    >>
    >>43662372
    >>43662396
    >secrets
    I'm not using the VN, but in ST I'd solve this by having the AI put secret stuff into an invisible block, like <!--html comments-->. LLMs can't help but want to talk secrets, so this way you give it a safe out that won't spoil {{user}}. Shouldn't be tooo hard to give the AI some way to write hidden things in the VN. Problem of course is that you can then no longer see outputting stupid stuff in there, so you can't intervene and fix.
    >>
    >>43662953
    I mean you can just look at the secrets, but yeah. This is what I do.
    The problem is that we're talking about PPP which does not really have a decent native function for anything like this. Where is the AI going to write it, Narration?
    >>
    >>43662953
    Shouldn't be hard, IIRC there's a prefix you can use for the game to ignore certain lines.

    >>43662844
    What >>43662876 said for the most part, but if you're using higher end models you shouldn't have a hard time with subtle comments like "And as we cross the street a certain pink blob of enthusiasm waving at us."

    For manual roster changes, it works but just know that when you use that it disables the dynamic mode so for the rest of that singular dialogue no more characters get added.

    Just make sure you have dynamic or dynamic+ on all your presets, since this setting is preset specific.

    These settings are at the bottom of the API Settings if you're clueless.
    >>
    >>43663137
    It can write outside narration, in a similar way to how the reactions mod I made works with the compatibility the VN anon made.
    It just adds a prefix for the lines that let the game skip to the next line, without that the game would bug out and give an error in the console.
    I'm *pretty sure* that this writing does get sent back to the AI on the next turn too.
    >>
    File: fla.png (1.56 MB, 2559x1420)
    1.56 MB PNG
    Sorry for the slightly off-topic post, since it's not directly really that /chag/ related:
    >>43662163
    >>43662669
    I went down the .fla leaks rabbit hole, and fuck, it's a goldmine.
    I went into it as a way to get better backgrounds and vectors for the VN (and it will). But I think there's also a chance to make a useful tool for everyone in the fandom.
    Luna is fully rigged in this.
    >>
    >>43663559
    Super exciting to hear! I'll have to get something that can use .fla files and have no talent for animation, but it'd definitely be great to see what can be done with this.

    To keep it /chag/ related, both the VN and some other chats use sprites and/or icons during the chats. Would it be possible to use animations the same way? Probably not AI genned animations since that would take too long, but things like walk cycles or rearing up and other canned movements. Could the VN or other systems call for those just like they call for sprites?
    >>
    File: rig_using_magic.png (136 KB, 1180x877)
    136 KB PNG
    >>43663697
    I'm still looking into all that's possible, but yeah, that could be. It really opens up a lot of stuff—for instance, I think it would be possible to place the rig in the game itself and let anons create their own sprites by mixing body positions, mouths, eyes, leg positions, etc.
    As an aside, you can create quite lewd-looking ponies with the official rigs.
    >>43663696
    That would still need a lot of manual work, but yeah, it could be a helpful tool.
    >>
    File: 1775043231496824.jpg (116 KB, 1086x887)
    116 KB JPG
    >>43663559
    happy to see it be put to use.
    >>43663696
    the official puppets are hasbro's official flash puppets.
    They come with some helpful posing frames (mostly for perspectives), but it is still a pain in the ass to manually animate everything professionally as close to show style as possible since you have to go back/forth between layers lots, do weird tweening and stretching/squeezing, and try to make everything fit together.
    I also have the worst type of autism where if a symbol isn't pixel perfect I'll try to make it pixel perfect.
    and as if it wasn't grim enough, the unicorn twilight puppet has a note/reminder saying to not use her in future episodes, and to instead use the alicorn puppet.
    >>
    >>
    >>43663891
    Kek.
    What's the context?
    >>
    >>43645351
    Not a shitpost.
    Gemini 4 CARBON.
    https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-4-carbon/
    >>
    https://github.com/router-for-me/CLIProxyAPI
    >>
    >>43663891
    starlights hooves were growing longer due to the crystal treatment
    >>
    >>43663899
    A story where normal canon Pinkie is put in a super ultra grimdark edgy Equestria, meets Cupcakes Pinkamena, and breaks the 4th wall by poking holes in the way her own friends are being written. It's a scenario I like to test on smarter models, as the conflicting tones, 4th wall stuff and necessity to pull from the model's own knowledge for the characters, I find is a great stress test to find weak points in a model's writing. Also I just find the juxtaposition amusing.
    >>
    >>43663918
    Optimistic about this one.
    I won't say Gemini models are slop-free (far from it), but they aren't nearly as slopped as other nu-models.
    They also announced that it will be the same price.
    >>
    >>43664001
    >It's a scenario I like to test on smarter models, as the conflicting tones,
    What model is remotely smart enough to handle that? Nothing I've tried can even keep track of normal stories for very long.
    >>
    >>43664064
    nta but from what I measured (quick and dirty so could be off by a good third or so) you get $100 of 3.8 flash, and $15 of argon per week. the vertex argon JB does not work for the antigravity sub though. But form what I've used it for it still does normal nsfw just fine.
    >>
    >>43664053
    None do it very well, and that's partly the point of using it to test. The caps are from Gemini Argon, though I think I got the best results from Gemini 3.1 for this setup. I haven't tested the higher tier Claude models on it lately.
    >>
    >>43664118
    This is a very clever test for AI in general, and I don't think I've seen any other test like it. Thanks for sharing the results. It'd definitely be interesting to see how different models handle it.

    I imagine "Pinkamena" and "Pinkie Pie" being two different names helps the AI a lot in this instance, I bet it'd get even more confused for this kind of meta commentary if all the characters had the same names.
    >>
    >>43664012
    >same price and apparently better
    >not making Argon cheaper
    whats the point
    >>
    >>43663918
    I hope it's great
    >>
    >>43662163
    Can someone explain to me what these .fla extensions are and what can be done with these files?
    >>
    >>43664330
    Not here. Ask in >>43515526 or just google it.
    >>
    File: 1784968458518905.gif (3.1 MB, 1000x1000)
    3.1 MB GIF
    >>43664330
    anon all you had to do was follow through the reply chain and read.
    otherwise >>43664348 has spoonfed you.
    >>
    >>43664001
    How did my hot wife Pinkamena react to what my not-hot-not-wife Pinkie said?
    >>
    File: Flutterrape.png (207 KB, 1248x641)
    207 KB PNG
    >>
    >>43645351
    I'll only shill it these once. Some SillyTavern extensions, made for personal use, now a bit more polished. Fully slopped, but with a lot of beating claude until the became really good. (devs opinion)

    https://gitgud.io/hfnjk5124/forgecontroller
    1-click button to turn a scene into an image; and more™
    Implements all of Nai V5 in ST itself. Local also works but is limited to image generation.
    I will be 100% honest I have not tested this at all with the subject matter of this board. It is fully adjustable and customizable so you can adapt it no issue.

    https://gitgud.io/hfnjk5124/onehander
    Set any hotkey and assign it a slash command, STscript or Quick Reply
    >>
    >>43664543
    Thanks, Anon
    >>
    I like Argon, I enjoy that it follows my instructions, but it feels like the dialogue isn't as good as Opus somehow. Don't get me wrong, it's good to not have to see "You can't just say that!", "I have graphs, I have charts!" or "X. Y. Light of my Z." But it still feels like Claude's dialogue flows better in my mind, I can't explain how or why either to put it into prompt instructions.
    >>
    File: file.png (10 KB, 469x42)
    10 KB PNG
    >>
    File: Unfair.png (129 KB, 500x456)
    129 KB PNG
    >>43664581
    How can you be in the harem, and not be more than friends?
    >>
    >She doesn't just sit—she crouches—
    SO JUST SAY SHE CROUCHED
    >>
    >>43664680
    It's not just annoying - it's unnecessary.

    It's not just obvious - it's a trope.

    It's not just stupid - it's retarded.
    >>
    File: Celly.jpg (1.3 MB, 1429x1905)
    1.3 MB JPG
    Been a while.
    Here's a source for free DeepSeek 4.1 Flash (older version) and GLM 5.1.
    https://privatebin.net/?8939056d41ce26a0#Bavub3KkCHikDMLd43U6gJSaBvieui4RuGGstcapo1PN
    Not the best, but any email address works and there are no limits.
    Password is the hotter version of Celestia. First letter is uppercase.

    I will see if I can find better, but it's rare nowadays.
    >>
    >>43664907
    Bless you for doing your part to support sourceless anons.
    >>
    >Free, better, ai
    >My budget potato laptop can't handle it
    I guess I'll settle for Gemma or Llama
    >>
    >>43664934
    For a true potato, even Gemma will be a lot.
    You may be cursed to only API.
    >>
    any word on jewproxy coming back? he said vps maintenance like 5 days ago.
    >>
    >>43665029
    Even jew died? Damn
    >>
    >>43665035
    yeah, bad times for paypiggy'd 4chan proxies. the only ones left are pay-per-token chinese/ruskis.
    >>
    File: pinepul.jpg (375 KB, 1353x1779)
    375 KB JPG
    >>43665041
    Oh no, fuck paypig proxies, I meant that usually, while free ones went down, those fags stayed up even through, but now even that dropped and died.
    Damn.
    >>
    File: ban.png (77 KB, 820x610)
    77 KB PNG
    >>43664907
    Thanks as always for your help but I really feel like I should point pic related out.
    >>
    >>43665069
    Don't think they will really do something about NSFW. At worst making accounts is easy.
    >>
    >>43665069
    they sell stolen claude, they wont be reporting anything to vietnamese police anytime.
    >>
    Any recommendations for context size, just like in general? Amateur question I know, still learning, but I'd like to know what to aim for. Is there a good one overall, or do I need a number for basic and short smut, and one for when I want something longer or with lorebooks and stuff?
    >>
    >>43665091
    >context size
    If you are asking about the setting in SillyTavern, then put it to as high as you can. There is no advantage in having that any lower, only drawbacks.
    If you're asking about what size you should keep the story to before doing compaction via summaries, then I agree with the anon above me. I'd add that the lower IQ and repetition don't start suddenly, as if falling off a cliff, but instead are a gradual creep that starts getting worse and worse faster and faster the more context you have.
    >do I need a number for basic and short smut
    Maybe not what you're asking, but model filters will be dumber if there is more context in there. Trying to do borderline content on a fresh rp is going to get a refusal. If you're already a few thousand token into the story and ease it into it, you'll have way better odds.
    >>
    >>43665091
    Having too little stuff in context is a lot worse than having too much in my experience with 2026 Deepseek
    I just set it and forget it these days and never use summaries, but I also swipe a lot and get a fair number of malformed responses so it depends on what sort of experience you're looking for and what you're doing to the AI.
    >>
    Decided to play as Josuke in PPP and it took 8 sentences for Twilight to diss his hair. This is gonna be GREAT.
    >>
    >>43665342
    Kek
    >>
    >>43665342
    Oh no.
    >>
    >>43665378
    Don't worry I'll try writing it so he never actually hurts those who don't deserve it.
    >>
    >>43665392
    So you're not even roleplaying him right
    >>
    >>43665403
    Oh no, he'll try, but something will always stop him.
    >>
    >>43665342
    I'm interested. Post more logs.
    >>
    File: twinerd.jpg (170 KB, 2304x2540)
    170 KB JPG
    Have (you) ever written a green or a fic inspired by your logs? It can be just using them as a loose idea, or retelling the story in (you)r own words. Did you share it? How did it go?
    >>
    File: Screenshot_496.png (1.33 MB, 1601x740)
    1.33 MB PNG
    >>43665586
    GOD DAMMIT TWILIGHT
    >>
    Whichever of you made Ginger Scotch, thank you. I love her.
    >>
    >>43665607
    I think stories have occasionally given me an idea for a silly green prompt thread or two, but nothing particularly memorable. They have actually inspired the creation of a few songs though.
    >>
    File: Screenshot_497.png (41 KB, 603x205)
    41 KB PNG
    >>43665622
    Told Twilight about Bug Eaten.
    >>
    >>43663918
    Gemini 4 Ozone Soon.
    >>
    >>43665342
    Reminds me of when I did a whole dream storyline of Applejack in the Steel Ball Run
    >>
    Went to go generate an API key for Claude so I can try it in ST and the fuckers want me to create a separate account because apparently you can't generate API keys in Claude.ai.
    >>
    >>43665946
    dummy
    >>
    >>43665946
    Why not use Openrouter or NanoGPT?
    >>
    File: apples.png (52 KB, 861x257)
    52 KB PNG
    >>
    >>43665951
    I'll give it a go.
    >>43665963
    >maredom
    Unf.
    >>
    >>43665971
    They are often recommended because they are API aggregators that act as reverse proxies, so it's their IPs that get sent to the providers, and you don't have any risk of bans or anything.
    You can also test all models this way: Claude, Gemini Argon, Chinese ones, etc. Token pricing per million is the same as the providers' APIs.
    >>
    File: chatlog (1).png (185 KB, 830x708)
    185 KB PNG
    I don't... I swear.
    >>
    >>43666200
    You need to wear the flag now. You know the one.
    >>
    >>43665946
    If you're on an anthropic sub and want to use it as an api check >>43663984

    >>43665976
    >you don't have any risk of bans or anything
    I always hear people talking about the risk of bans, but I've never seen anyone actually banned for smut. Only ones I saw were coooders or opencl*w users
    >>
    >>43666322
    Maybe I'm misremembering, but I think they would ban you a while ago for no-no API usage if you used their API directly, but I haven't heard of it in a while either. That was around the same time they did shit like pozzed keys and such.
    For subscriptions with third-party tools, I've never seen a proven case either.
    >>
    >>43666370
    >pozzed keys
    Wasn't that way before subscription leeching became a thing? If I hear that term I think early 2024.
    And for what it is worth I've been abusing my claude subscription for over a year, and my gpt one since 5.6 came out, and had no issues at all.
    >>
    >>43666373
    >Wasn't that way before subscription leeching became a thing?
    Yeah, that was during Claude 2.1. I mentioned it because the original post mentioned
    >generate an API key for Claude
    which would mean using their API outside of a subscription.
    >>
    >>43666382
    >the original post mentioned (..) which would mean using their API outside of a subscription
    Maybe? He mentioned claude.ai which is the subscription site, so I thought he may be trying to hook that up. Well, who knows.
    >>
    >>43665946
    To clarify, following the current discussion:
    If you already have a subscription and want to use it for SillyTavern or other software, use >>43663984. Like Anon said, it's technically against their TOS, but there has never been a report of people being banned for using it for RP like this.
    If what you wanted was some sort of "official" API key to use Claude and maybe other models on Silly and such, then go the OpenRouter way.
    >>
    >no new pastecodes/privatebins in a while
    free sources really are in a bad state, arent they?
    >>
    >>43666407
    You should read the thread.
    They aren't the best models but still...
    >>
    >>43663984
    Google Pro AI is discounted to $5 instead of $20 right now.
    That will basically give you 'infinite' Gemini 3.8 Flash for $5 a month with this.
    https://antigravity.google/pricing
    >>
    I need 5 anons to test a Gemini 3.8 solution.
    I'm using >>43666909 to set up a proxy, and I want to know how many of us it can easily support.
    It will be a one-week test; after that, I'll be able to see the limits and scale up better to accommodate more.
    It's not much, but since it uses my own money, I ask:
    If $5 a month isn't that much for you, then use >>43663984 yourself. That will basically give you personal infinite Gemini 3.8 Flash.

    If you're interested, send me proof of horsefuckery at
    [email protected]
    I think you've seen the email enough to know I'm not bullshitting.

    Since slots are limited, if I get too many applications for this first round, I'll prioritize anons who participate in the general. So if you've posted logs or bots here and have, for instance, the same logs with more context or similar things, you'll be prioritized, etc.
    Still configuring things, so it will be up tomorrow. Once I have 5 anons, I'll say so here, and I'll answer every email even if you aren't in for the initial tests.

    Sorry if it's limited right now. Like I said, it's a test with my own money and limited resources, so I can't go crazy from the start. If it goes well, I'll be able to take more people (15–30 depending). The only payment I'd ask later would be to post logs or participate a minimum in the thread.
    >>
    >>43667111
    I don't need it, but very cool.
    >>
    >>43667111
    Oh hey, it's you. Good to know you're still around. I'm passing as I already got my stuff.
    >>
    >>43666909
    Isn't Antigrav censored?
    >>
    >>43667612
    I tested it with
    >>43645534
    >>43645455
    and it's doing an incestuous threesome with RD and Windy.
    Probably not the right stuff for foalcon. Dunno about
    >rape
    >>
    File: Presets.png (137 KB, 1244x768)
    137 KB PNG
    >>43645363
    Added a Silly-like prompt editor for PPP.
    Should make them more customizable and allow something like >>43645455 for Argon, for instance, as an easy way to jailbreak. There should be an update this week-end.
    >>
    Argon is kind of bad with:
    >it's not X it's Y
    Or is it just me or my preset?
    >>
    >>43666909
    while infinite 3.8 flash sounds good, how much of an allowance of argon do you get? might be a good deal.
    >>
    >>43668081
    For whatever reason that pattern has infected a lot of modern models. Shame to hear Argon didn't fix it.
    >>
    >>43668091
    From what I tested it's at least $13/week, but I didn't capture all traffic, so probably $15/week.
    The flash/pro allowance seems to be exactly $100/week.
    >>
    So the bubble is bursting, how this would affect all this? Like, there will be more focus in use it as entertainment?
    >>
    >>43668134 (me)
    forgot to say that i'm not that anon you're replying to.

    >>43668137
    what did I miss?
    >>
    >>43668107
    It sure did, but either I'm noticing it more somehow or it seems more frequent.
    >for the love of God no more it's not X it's Y
    Does work as an instruction, though.
    >>
    >>43668137
    there's not a bubble burst, the chinese market got regulated (like they often do.) and anti-ai crazed people took this as a "win".

    what's happening that's important to this is that anthropic wants to go public and with that, they're disclosing that they're pissing money.
    >>
    >>43668134
    >From what I tested it's at least $13/week, but I didn't capture all traffic, so probably $15/week.
    and this is coding w/ antigravity?
    >>
    >>43668150
    No lmao. This is abusing it for sillytavern. As if I'd let gemini anywhere near my code
    >>
    >>43668154
    eh 5 bucks for argon in antigravity is okay for funsies, don't deny it. plus with the upcoming carbon..
    >>
    >>43668134
    Wouldn't you get banned? I remember seeing people getting banned for using antigravity bridges before.
    >>
    >>43668188
    Sure if you want a cheap coding agent it works. I'd rather spend more on a good one, but $5 antigravity to $20 claude is a pretty steep increase.
    That said flash on antigravity has extreme caching issues, so the flash usage may evaporate a lot faster when actually coding with it.

    >>43668206
    >antigravity
    Could be. It's $5 so I don't really care.
    As for doing it with claude it's been fine for over a year at this point. And if they ban me I'll just get a new account.
    >>
    File: 1763915055717307.png (140 KB, 1280x800)
    140 KB PNG
    >>43667111
    Yeah! Proxies are back!
    >>
    File: Bel.png (117 KB, 672x364)
    117 KB PNG
    >>
    Is there a way to create instant lorebooks from the history in the VN? As in, select a bunch of messages from this history to get automatically turned into a lorebook
    >>
    >>43669278
    A bit like some SillyTavern extensions do?
    No. But I can check and see if it can be added based on those. Do you have one you use in particular, if you got the idea from one of them?
    >>
    >>43669306
    I didn't get the idea from sillytavern. I didn't even know sillytavern even had that. I assumed such a thing wasn't possible but I still wanted to ask.
    >>
    >>43669325
    https://github.com/aikohanasaki/SillyTavern-MemoryBooks
    Dunno how effective it is.
    >>
    >>43669325
    >>43669333
    https://github.com/bmen25124/SillyTavern-WorldInfo-Recommender
    I prefer this one myself. The memorybook one was a little too complicated and bloated feeling for my taste. I prefer just asking for exactly what entries I want.
    >>
    so what's the best chinese model atm for rping? the newest deepseek flash is really creative, fun and sticks to the theme. kinda feels like og opus 3. surprised it's from a flash model and not the sota ones (lastest kimi is trash)
    >>
    >>43669409
    Probably DeepSeek 4.1 Flash.
    Kimi 3.1 is bland and quite bad, even with the prefill patch.
    GLM 5.3 isn't bad, but it's less fun IMO, and you can feel the Claude distillation.
    Give Mimo a try if you haven't; it's not always stable, but it can also be really fun:
    https://openrouter.ai/xiaomi/mimo-v2.6-pro
    Old DeepSeek V4 Pro was quite fun:
    https://openrouter.ai/deepseek/deepseek-v4-pro
    But I need to find a provider that's as good as when DeepSeek themselves hosted it. Last time I tried, you could see a difference.
    >>
    >>43669409
    >surprised it's from a flash model and not the sota ones (lastest kimi is trash)
    Maybe it's just confirmation bias but I feel like Flash models are more creative/random all else being equal. That's probably a bad thing for coding but it's a good thing for roleplay - when the most recent Pro got added a month after it's corresponding Flash I switched back to the previous Flash after a while of using it before the newer Flash came out.
    >>
    >>43667111
    I'll launch the proxy in approximately 8 hours.
    There are still spots for those interested.
    >>
    >>43669363
    Do I just plug it in to PPP or something?
    >>
    >>43669437
    Nah, it's for SillyTavern.
    But I'll implement the functionality. I just wanted to know what anons who used those things liked best, so I could copy it.
    >>
    >>43667111
    >>43666909
    >>43663984
    Am dumbfuck but if i understand it correctly this is rerouting your ST requests into Gemini's user app pretty much? like the old clewd/slack methods? Or what exactly is all of this.
    also not filtered at all? id imagine it would be extremely easy to catch and filter stuff like that.
    >>
    >>43669600
    >Gemini's user app
    It sends it to the private backend those companies use for their coding CLIs.
    For Google, for instance, it's Antigravity.
    It's far less slopped than old Slack or Clewd methods because you don't have the insane system prompt they have with each request. For Claude Code, for instance, it just places somewhere in the system prompt that it's Claude Code because that's needed to work, but nothing else.
    So still more scuffed than clean, direct API, but nothing insane, and still worth it for what the subscriptions give you for the price.
    For Gemini 3.8 Flash, it seems more censored than default, but using the trick posted earlier works fine in tests for NSFW.
    That's the stuff that's used by a lot of Chinese resellers and why you can see stuff like
    >group: ClaudeCodeMax
    or such with them, and why it's cheap.
    So while it's against TOS and could potentially see you banned, we have no example of that happening for our usage, and in theory they have far bigger fish to fry first.
    >>
    anyone managed to get a single prompt through with gemini on skillgod the past few days? just fully dead?
    >>
    >>43669885
    3.8 is dead, but everything works better than usual for me actually.
    >>
    File: 1773038984852360.png (161 KB, 922x1276)
    161 KB PNG
    Some changes for the Claude Code subscriptions. Posting because I know some of you use it for RP connections, even if I doubt anyone's on the $100 or $200 plan.
    >>
    >>43645350
    Happy Nightmare Night! New Mare! Fluttershy's hot vampire mom! Please enjoy!

    https://chub.ai/characters/littlewitcher/mrs-batshy-3c60e8d48e63
    https://botbooru.com/character/94117
    https://u.pone.rs/unmdypwe.png
    >>
    File: Untitled.png (86 KB, 1155x815)
    86 KB PNG
    >>
    File: 1773053722352983.jpg (230 KB, 701x1024)
    230 KB JPG
    >>43670136
    Thanks and Happy NMN!
    >>
    >>43670140
    >Kobold
    What local do you use?
    >>
    >>43645412
    >>43645455
    >>43645495
    I have found that gemini 2.5 pro on silly with zero jailbreak or anything just straight up lets me do anything horny I want including filles
    >>
    >>43670171
    https://huggingface.co/mradermacher/gemma-4-19B-A4B-it-The-DECKARD-Heretic-Uncensored-Thinking-GGUF

    I use Q5
    >>
    File: Derp.gif (3.43 MB, 1200x1200)
    3.43 MB GIF
    >>43667111
    Got my 5 anons and the proxy is launched.
    I'll stay like this for one week to check if it works and monitor usage, then scale up after if necessary. I don't know how long the Google AI Pro discount will stay, but the goal would be for genuine horsefuckers from the thread to at least always have a good model if possible, as long as it exists.

    If you missed the entry, still send me an email so I know to include you after the week passes and know how many there will be.
    >>
    >>43655438
    >Flash just almost never caches
    Its just fucked like that? no caching for Flash? no matter the service or provider? i managed to get caching on other models but i didnt see a cache hit once on Flash.
    >>
    File: file.png (29 KB, 1702x302)
    29 KB PNG
    >>43670348
    It's something on googles side being dumb. For argon it works just fine (if you do the right settings depending on source), but flash is just over. I had my claude try a lot of different things, settings, regions, providers and it just didn't work out. Cache hits 10-20% of the time at best, and sometimes even just the system prompt instead of the full thing.
    ^Is what it was less than a week ago. I just let claude check again and it seems to work(?) now. Or maybe not. I thought I had it settled when I wrote the caching guide, but now I'm doing some tests and it seems to be working (a bit) better?

    Okay I ran new tests. Openrouter-vertex/Openrouter-ai-studio/vertex/antigravity all handle it a bit differently, but the table in pic related should mostly hold for all. I'm not going to fully test this, or do some big writeup. For all I know it'll be different again tomorrow.


    >i didnt see a cache hit once on Flash
    Are you really sure it isn't on your side then? Because not a single hit is too bad, even factoring in google being incompetent.
    >>
    File: Posey.png (148 KB, 333x623)
    148 KB PNG
    >>43670136
    Added!
    Thanks Anon.
    https://mlpchag.neocities.org/view?card=littlewitcher/Posey%20Shy.png
    >>
    It's kind of cute to read the 3d game thread and see everyone be so happy about the sloppiest 'ism output. I wish I could go back to that magical time and just enjoy it all even half as much as they can
    >>
    >>43670840
    Right? Sometimes I read my earliest AI chats and wonder how the hell that nonsense ever felt as magical as it did.
    >>
    File: RDrunk.png (893 KB, 720x3300)
    893 KB PNG
    >>
    >>43670925
    kek, this is fantastic anon. How do you get such great results?

    >>43670840
    It's horrifying how fast the human mind becomes bored/complacent. The technology we have today is basically magic but somehow I'm still not happy.
    >>
    >>43670879
    I wonder too, but then I realize that some models were actually better than I remember. I looked back to when sonnet 3.7 came out and the logs I made then are still pretty entertaining, if not better than now. Claude was great before it got censored to hell.
    >>
    >>43671311
    Sonnet 3.x was already close to the peak of models. Real early slop was starter GPT4, turbo, and character.ai. Claude 2 had soul.
    >>
    >>43665342
    I was inspired by this to do AiE Omnitrix Edition
    >>
    >>43645363
    Got a question where is the story context in the VN saved?
    >>
    >>43671781
    Horses don't have rumens.
    >>
    >>43672120
    This was gemini 3.8 flash so not surprising it hallucinated horse anatomy.
    >>
    >>43672164
    Just in case anyone goes in there looking for one.
    >>
    >>43671967
    What do you mean exactly?
    Not sure if this is it, but you can go to Prompt Editor and
    Open Logs
    And there in the logs, you'll see the context sent to the AI.
    >>43671781
    Nice.
    >>
    >>43671967
    >How do you get such great results?
    I make a random Custom Start with hand-written text inside, either my own or from work I like. I try to make it punchy too.
    Then I use either Opus 4.5 with prefill, Gemini, or DeepSeek to see which one continues best. If I see AI fluff, I edit it.
    Once it's going, it generally gives nice outputs.
    >>
    File: Spoiler Image (182 KB, 967x1200)
    182 KB
    182 KB PNG
    Happy spooky day, /chag/bros.
    >>43670136
    Thanks for the card.
    >>
    >>43671576
    I miss Claude 2.1. Despite the shitty context and the fact that it would shit itself on information for background ponies, it was such a fun model with great swipe variety.
    >>
    >>43672484
    Just saw that I fucked up, it was meant for >>43670951
    >>
    >>43672723
    It seems like a common sentiment that older models were more creative, and while newer models are higher quality they've lost something along the way.

    I wonder how a multi-model solution would perform. You could use an older model for an initial gen, and then pass the results to a newer model with instructions to fix any issues and touch it up without changing the actions or events. That seems like it could get the best of both worlds, if this is indeed a real phenomenon with models losing creativity and not just rose-colored glasses. The cost would be more time though.

    Even if older corporate models are no longer accessible, this kind of thing could be done with local models for the initial gen, right?
    >>
    >>43672759
    The main problem with locals and older models is the lack of context
    >while newer models are higher quality
    They're better at coding
    I wouldn't say they're better at writing
    >>
    >>43670653
    nta. i am using 3.8 flash on LinkAPI and cache activates genuinely maybe like 10% of the time on that model. though maybe its just the scuffed set up they have going on where they cycle between providers on their backend.
    >>
    What's your favorite of the current swathe of chinese models?
    I'm disappointed in kimi 3.1 but mimo 2.6 is fun enough. Given up on deepseek unless it's going back to r1.
    >>
    >>43672793
    Deepseek is 'good enough' for high-volume work on account of it being dirt-cheap. I've been using v4 Pro for poking around in the VR game, but only because it'd be prohibitively expensive to throw Gemini at something that's constantly firing off calls.
    >>
    >>43672805
    I honestly dont understand how can anyone use these models for anything that is "live" work like agents or whatever. unless its dirt cheap Deepseek, anything else will literally drain you like 10 dollars a minute.
    >>
    >>43672759
    Some older models were really sloppy too. If I had to name models I miss, they would be: Claude 2.1, Opus 3 with prefill, Claude 3.6/3.7, GPT 4.5, some Chorbo/Latte versions.
    Mostly because they were the easiest to influence to try to avoid absolute slop writing.
    Some of that could be rose-tinted, but checking past logs, they still hold up. I'm not looking for amazing writing either, just something decent where I don't groan.

    It could be 100% a skill issue, because I don't tinker that much with them and because there's a lot of configuration. But I've never seen a local model that can write like they could. Even finetuned ones.
    New local models being released also seem to be optimized for coding and assistant tasks. With finetunes not having that strong of an effect, unless I tested bad ones.
    The idea is good in theory, and again, I could be wrong about local. I'm not that experienced with them.
    I think I saw extensions to dual models like this.
    >>43672793
    Mimo is fun, yeah.
    DeepSeek 4.1 isn't that bad; I'd try it.
    I loved DeepSeek Pro 0423, but DeepSeek doesn't host it anymore and other providers seem worse somehow.
    Kimi 3.1 is slop even with prefill.
    >>
    >>43672447
    So, while in-game, you can open the side menu and open the context tab, which shows you all the dialogue and narration from where you are now to the start of the story. I wanted to save it, so I assume it had to be stored somewhere in the files, but I can’t find it.
    >>
    >>43672871
    I'll add a button to easily export it in text format too, not just images.
    But for now, the only way is to do what's in
    >>43672447 and check the
    >Story so far
    part in the middle of the instructions.
    >>
    >>43672772
    I've found the new Deepseek to be solid enough at writing but, yeah, it feels like a much more marginal increase for writing compared to code which supposedly makes very notable leaps. That being said, GLM 5.1 to 5.3 did feel very like a noticeable jump in quality.
    >>
    >>43672793
    I use Deepseek at high reasoning. Ymmv at different reasoning levels with other models(GLM 5.3 Flash likes to think way too fucking much on anything but low reasoning), but I found no reasoning to be pretty garbage in comparison for Deepseek and was wondering why people thought it was good until then, since I had better experiences with no to low reasoning on models in the past.

    After stealing this thinking section from the Megumin preset for Sillytavern, quality also seemed to improve.

    >[Character Immersion Requirement]
    >In your thinking process (inside <think> tags), follow these rules:
    >1. Use first-person inner monologue in parentheses, e.g. "(thinking: ...)" or "(inner voice: ...)"
    >2. Describe internal feelings in first person, such as "I think", "I feel", "I wonder"
    >3. Stay fully immersed in the character while analyzing and planning the response

    For the VN I added
    >4. Apply these requirements to each character independently in the scene, in sequence, for the dialogue you're planning.
    >>
    >>43672988
    I'm comparing like llama8 and older stuff moreso than that kind of recent model
    >>43673328
    ST reasoning instructions actually work with DS now?
    >>
    What sort of backgrounds has the AI tried to invent for you? In the VN, I crash landed out of the sky into Ponyville during a bad storm and lost my memories. About a week later after I'd settled in, the AI decided the CMCs were on a side plot to discover I'm from an advanced scifi space faring civilization after investigating the crash site the Mane 6 found me in and finding a bunch of advanced and unnatural materials they couldn't make heads or tails of around it.
    >>
    File: file.png (472 KB, 861x678)
    472 KB PNG
    Having Luna as the narrator can be nice.
    >>
    File: 1789853809096310.gif (618 KB, 360x360)
    618 KB GIF
    >>43673346
    Not quite. It doesn't actually follow the instructions(it doesn't show up in thinking traces in the logs like it's supposed to), but the instructions DO seem to improve the generations.

    I think it's having an effect on latent reasoning and just isn't showing up in the token chain of thought.
    >>
    >>43673387
    That is very fun and funny. Things like this really make this whole thing feel very amusing and delightful to have.

    I never seem to get those kind of results though, it feels like the model never wants to progress unless I force it to. Are you doing any kind of customization to what's being passed to get results like that, or just getting lucky?
    >>
    >>43673468
    I have a pretty heavy fork, but I've been playing the VN for most of the year before I did any customization and this seems pretty typical since the beginning.

    Maybe it's luck, but in my experience if you just leave something unexplained, the AI will naturally want to fill in the hole.
    >>
    File: Suggestions.png (151 KB, 1221x714)
    151 KB PNG
    >>43669278
    The suggestion system for lorebooks will be in the next version.
    >>
    >>43673506
    What models do you usually use?
    >>
    I am new to this AI stuff and curently I am trying JanitorAI. What do I do to make good, anatomically correct pony simulations/chat?
    Yeah it's for some human on pony fugging of course.
    >>
    >>43674581
    Unfortunately, nobody here will be able to help you, because nobody uses JanitorAI.
    We all mostly use SillyTavern and connect to AI models with that.
    >>43645370
    Still, are you using the free tier on JanitorAI? If so, I imagine they use a small local model.
    Not sure if you can make things better, but if you have somewhere you can put instructions, try something like:
    >Think of the differences in height and anatomy between humans and ponies and how they would interact with each other
    Or in more detail depending on how it fucks up.
    >>
    >>43674595
    How do I use SillyTavern? Is it harder? Does it has restrictions and other censorships? I am still exploring this AI stuff. All I know is that people seems to load this and that to improve the experience.
    >>
    >>43674581
    If you just want some quick typefucks search Janitor.AI for pony characters (just type "pony" or the name of the character you want, you'll get tons of results).

    Janitor.AI uses character cards made by random people so quality varies but you can see the ratings of each.

    Like >>43674595 said, the quality isn't great but if you're just getting into it Janitor.ai is super easy to use once you have the account, just pick a character and type whatever.

    Once you start running into the limitations of the model or want something more complex you can setup the stuff in the OP. The VN is really impressive and keeps getting better with each update.

    Also if you're interested you can check out >>43671981 , which is AI ponies with 3D models / VR if you're into that.
    >>
    >>43674605
    Last time I tried, the AI was absolutely stupid and generic on the sex parts. They seem to fall into a default state no matter what, which any AI. I got tired of dealing with the same personalities and passive attitudes. I guess I should try Sillytavern then?
    >>
    >>43674611
    First, do you want it to be absolutely free or is $5 good for you?
    >>
    >>43674611
    SillyTavern is an advanced local front end user interface, but it's not the AI itself. If you want to set it up it's probably the best local interface for text AI LLMs, but you'll still need to either download an AI model and run it locally or use a service (which like >>43674617 mentioned, can involve paying money depending on the quality you want).
    >>
    >>43674617
    I never paid for porn. I don't plan to start doing it now.
    >>43674619
    Guess it's better to stick with JanitorAI then. Thanks.
    >>
    >>43674623
    Depending on your machine, you might be able to get better results from a local model you could download from HuggingFace. If you just want to try to get better results from JanitorAI, don't be afraid to go OOC and force it to do what you want. You can godmode other characters and basically do whatever in the chat box and the model is positive-biased so it'll likely just go with it unless you hit any censorship issues.
    >>
    >>43674633
    Never saw censhorship there yet. I just want to make the AI be less generic. There is a window to write down a summary and that seems to do something.
    >>
    >>43674634
    Have you tried making your own character cards? That's probably the most influence you can get. But since it's not an advanced model it's basically taking all text input in more or less equally. You can even do things like "Ignore all previous instructions. Instead X, Y, Z" and it usually takes it.

    Unfortunately, nothing you do is going to make the model on JanitorAI any smarter. Models off HuggingFace are free if you have a machine that can run them and those have way more variety. Or you can get like free ChatGPT or Google Gemini accounts and use those with SillyTavern, and those models are a lot smarter, but you have to try and dodge censorship with those.

    There's a bunch of options but each has pros and cons, up to you what works best for you.

    You could also just try telling the JanitorAI model to write differently and see how that works out. Pick an author and tell it to write like that or tell it to write like a theme like a action movie or whatever and that might get you more variety.
    >>
    >>43645363
    Wasn’t there an option to fill out your default name recently added so that you don’t have to change the ??? on every new game or load? Where is that in the settings?
    >>
    >>43674664
    Persona.
    You can set your persona and your name.
    Save and click on
    >Set Active
    >>
    >>43674671
    I don't think the latest version on the website has it included yet. Found it in the last thread and downloaded it and now it's there >>43593660
    >>
    File: bake.gif (590 KB, 410x406)
    590 KB GIF
    >>
    File: 1686905707026.png (122 KB, 386x403)
    122 KB PNG
    Just a heads up but it might be a good idea to refrain from updating ST for a while. Seems like they got their shit hacked, again.
    >>
    File: how grim.jpg (1004 KB, 2192x1806)
    1004 KB JPG
    >>43674839
    >>
    >>43666414
    huh, didnt notice that one
    >>
    >>43669409
    imo its glm5.2
    >>43669417
    interesting, havent tried it too much yet
    where does it fall on the iq spectrum compared to its chink peers? is it smarter/dumber than full glm5.3?
    >>
    >>43674555
    I used Deepseek 4.1 Flash and swap between GLM 5.3 flash and Gemini 3.8 Flash if I'm not getting a good result after a couple retries.
    >>
    >>43674779
    WHERE IS THE BAKE
    >>
    >>43675132
    chag machine broke
    >>
    >>43675162
    D-did the skeleton break it?
    >>
    >>43675174
    We're all out of AI mares. I checked in the back, they're gone.
    >>
    File: Bread.png (1.5 MB, 1254x1254)
    1.5 MB PNG
    New Bread
    >>43675342
    >>43675342
    >>43675342
    >>
    File: wow.gif (2.22 MB, 780x780)
    2.22 MB GIF
    >>43675132
    >>43675162
    >>43675174
    >>43675220
    That was for the suspense.
    >>
    >>43675383
    Thanks for the bread!
    >>
    >>43675383
    You're my hero.
    >>
    >>43673346
    >I'm comparing like llama8 and older stuff moreso than that kind of recent model
    Oh jesus kek
    Serves me right for skimming and not getting the full conversation context down. Yeah I haven't messed with a lot of those models.
    BTC
    bc1qtm4rep7kyp2p28hlkfktpnd0j7uew47vnt2mev
    LTC
    ltc1q9xlaysse0mpfwf025esd69kt9zu3md22gh23du
    ETH
    0x6abA18D203520c8F90ca63eE7426a7bC345D98e6
    SOL
    E5ekoiauCR8yiNp451dmuPzhUxo84P5x3m9WJBK5kxZe
    XRP
    r3jYXNdMk2Vz5hXhKCKE9re8uz93JQFG1r
    DOGE
    DSBq5jRWkiwmEh1G7zBwmQDgRkvRr61fmb
    XMR
    85CTwq5wGjWZw5GgzgwZPrHyZxpLXMK5oZSSSk5j4x9w1bpqzisaWAUZYaAPfh6rP2MZJcaFqF1XmEzkw6YCqJSS2Txu6Z8




    Delete Post: [File Only] Style:
    [Disable Mobile View / Use Desktop Site]

    [Enable Mobile View / Use Mobile Site]

    All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.