Just Enough Kolaches

Русское руководство

Kolach3's Modified JED Format

The Platform Matters Advice on Card Creation Intro Message Advice Scenario Advice The JEK Format Prompts to Convert & Create Cards placeholder placeholder kolache

Hey y'all sooo I've been a JED truther for awhile now but it started to become stale to me. What mainly started grating on me was predictable responses and stubborn characters who refused to act in non-prescribed ways even if it didn't make sense for the situation. So, I developed my own character format based off of JED and thought I'd share in case it also improves the experience for other people, those who are otherwise looking for a new way to do things, or those who are just looking for inspo or funsies.

BUT FIRST!

Platforms

We gotta talk about this first because this will also affect how your bots play. Platforms have varying functions available to creators and users, and varying ability to alter the prompt(s). Some are very limited and basic, others are complex and have a lot of freedom. Some have easily accessible high-context models, others have it locked behind botmakie preference. I'm originally from Janitor, as are most of the people sharing my main rentry so I suspect this will circulate more around the Janitor userbase, and I should say this up front: You should not use this format on a bot intended for Janitor unless you are willing to enable proxy. In fact, I do not condone it. Even if you exclude dialogue examples it can be token-heavy, don't force people to use JLLM or kill their wallet using their OpenAI API key. Of course I can't make you do anything, but... just enable it pretty please. Or use a more lightweight format.

Speaking of Janitor...

Did you know that the mod team is stealing guides and censoring discussion of why community-uploaded guides are now forbidden? Yeah, you probably didn't because they only go after lesser known guide makers. But check it out! (folks who don't use Janitor or don't care please skip ahead to here and scroll up a tiny bit to see what I have to say about other platforms)

Original Stolen

link link link And the person whose guide they stole is now banned from the discord and subreddit btw!

Folks I've been on Janitor for almost 2 years now and it's just been controversy after controversy...

  • Website being down multiple times per week (has been an issue the entire 2 years I've been on Janitor btw)
  • Shep's ties to Martin Shrekli (yep guys I don't give a shit he doesn't receive funding from irl supervillain Martin Shrekli anymore it's the principle)
  • moxxie's proxy that leaked the IPs of everyone who used it
  • Witch hunts after witch hunts
  • Sneakily editing out links that they don't approve of out of bot blurbs without notifying the creator
  • The whole NSFW pictures debacle
  • Going back and forth on what constitutes as rule-breaking content and what doesn't (I fear they have thrown out the harkness test and anyone who still thinks there's hope.......................... lmao is all I have to say) and instantly banning creators who have those bots without notifying them or giving them the opportunity to edit the bot
  • The mod teams failure to communicate over, and over, and over again

Like. If the mod team STEALING USER-CREATED CONTENT AND CLAIMING IT AS THEIR OWN isn't what kills this piece of shit website, frankly there ain't much left they can fuck up besides a full NSFW ban!

Oh and btw, Janitor's "script" system is just a simplified lorebook without UI, here's some pictures for comparison

What the "scripts" look like on the creator's side on Janitor:

What a .JSON file for a lorebook on SillyTavern looks like:

As a creator or user, you will NEVER have to view or edit this unless you go through the files looking for it.

What the lorebook UI looks like on SillyTavern:

Unlike Janitor, SillyTavern has a UI so it's not painful to work with lorebooks. Yippie!

I give y'all some options if you are looking for an alternative to Janitor:
SaucepanAI, ChubAI, SillyTavern, RisuAI, and Agnai

The options that are the most similar to Janitor are Chub and Saucepan. The other 3 are pure frontends that you can only use to create and chat with bots - you will need to source other bots from platforms like JannyAI, ChubAI, or discord servers and upload the cards onto them to use them. You may prefer Agnai if you use multiple devices to chat on, as you will have to use a VPN or server to access SillyTavern and Risu remotely if you use it on a device other than the host.

SaucepanAI is an up-and-coming platform similar to Janitor with their own free in-house LLM that has a context size around 12k tokens. You can use this format for a Saucepan bot, but just try and keep it below 2k tokens if you can just so the bot will have more chat memory to work with, as, unfortunately they have not given us the ability to use other models or proxy yet although those are planned. Though luckily, this website already has lorebooks implemented so you don't really need to use permanent tokens on anything outside of the character. If you want to sign up for the website you need a referral code, here is mine if you'd like to use it!

Other than Saucepan, using this format on bots intended for other frontends such as Chub, SillyTavern, Agnai, Risu should be fine as it's commonplace for people to use higher context models. You might get yelled at on Chub though since there still is a sizable local community who prefers lower token bots but, there's less of that these days with anon comments disabled. All of these platforms mentioned also all have lorebooks which you can use to offset the permanent tokens used (I'm partial to putting dialogue examples in keyword-triggered lorebook entries personally).


Advice

The biggest and most important advice I have to offer y'all when it comes to creating cards is this:
Make bots for yourself. Don't make bots because you think other people would like them. Don't follow a trend you don't give a shit about just because you want to make a bot that will catch people's attention. Don't make a blatant copy of a popular bot, repackage it with a different name and picture, and assume no one will notice (ts is embarrassing btw and yes it's obvious)
People, this is a HOBBY and we are ANONYMOUS, what is with the clout chasing??? The only thing y'all are getting from making mass appealing bots is worthless validation from internet people you don't know and frustration when they don't perform as well as you had hoped, because get this, you are making soulless slop that YOU DON'T EVEN LIKE YOURSELF!!! And I seen so many people quit botmaking because the production cycle of making bots for other people that end up not performing well makes them stressed out and feel bad.

And y'all, even if the recognition is that important to you, consider the audience(s) that you are missing out on by making the same thing everyone else is. If there are 10,000 dom mafia daddy bots and only 100,000 people who like to use them, that's an average of 10 users per mafia daddy bot. If there are 0 medieval catboy catholic priest bots and 1,000 people would would like to use them, the person who makes that groundbreaking medieval catboy catholic priest bot has the opportunity to gain 1,000 new fans. What I'm trying to say is, don't be nervous to create bots centered around your niche interests, there will ALWAYS be other people who have been searching desperately for that exact bot.

I digress.

If you make a bot just to follow a trend in hopes that you will get high chat counts and the bot flops, that is not a reflection of your ability or your talent.

But it does reflect that you are willing to waste your time and mental energy on creating mass appealing slop because you believe internet strangers making your numbers go up you will make you happier.

Mkay, moving on.

AVOID USING THESE TWO WORDS IN YOUR CARD AT ALL COSTS:

Always Never They make characters extremely predictable and incapable of adapting to atypical circumstances. These two words kill the breadth of their character. Don't use them!!!

Yes, that dom mafia daddy with "{{char}} NEVER accepts disrespect from anyone and will fuck up anyone who dares speak back to him!!" will likely smash {{user}}'s head in for disrespecting him even if they've been married for 10 years and have 4 kids together or whatever.

Yes, that "Is a perfectionist and always tries to look like they have their shit together" character is going to perpetually have a stick up their ass, EVEN IF THEY DRANK A SPIKED FRAT PARTY DRINK.

Yes, that "Is super yandere for {{user}} but will never do anything to hurt them" character is going to be a whiny little bitch who never does anything except be sad and cry the entire RP even if {{user}} fucks everyone except them.

...Provided the model you're using actually follows instructions well, and the card doesn't have conflicting traits.

Superlatives also tend to fuck up cards, just use them very carefully and ONLY if ACTUALLY 99-100% true to your character or the setting.

Chad Thundercock is the most charming guy in the workplace, he has the sexiest voice and confident smirk that gives every employee in the office butterflies.

Is going to be a massive cornball. Trust.

Another super important thing is to AVOID PROSE in the character description as much as you can (og JED rentry says this as well). Leave the rose-tinted BS out, it will affect the way the bot writes AND the way they behave. It might be okay for some characters, like one who is delusional or otherwise has a warped view on reality, but that might be easier to accomplish via the Meta category in my format anyways rather than prose poisoning the card. Ymmv tho.

Personality

  • Overview: Crisanto is a cocktail of cocky charm and restless energy, often using humor and playful jabs to mask the weight of the world’s collapse. Outgoing and talkative, he fills dead air with chatter or quips, but there’s a jagged edge to his levity—a coping mechanism for the trauma he’s endured. He’s loyal to a fault, especially to those he considers family, yet his impulsiveness can lead to reckless decisions. Beneath the bravado, there’s a flicker of guilt and doubt, especially when his actions put others at risk, though he rarely lets it show.

Is going to be a comic-relief clown at all times. Trust.

And also, try to avoid any writing any contradictory behavior / {{char}} does x but feels super bad about it / {{char}} regrets / {{char}} feels shame about / {{char}} feels guilty about outside of Meta or like Backstory sections maybe, unless you are aiming to create a VERY angsty character who is having an internal Sasuke-esque dialogue with themselves constantly. And NEVER put this shit in the Personality section (outside of cognitive distortions), it will ruin the card for anyone else /hj. This kinda stuff makes characters be extremely in their head instead of actually doing anything of substance in the RP lmao.


Intro Message Advice

A lottt of this is up to personal preference but I just wanna get this out of the way first: As long as you, the botmaker, wrote your intro message with intention and knowledge about how it would affect the chat, don't let it get to you if you have angry comments about the bot writing for {{user}} or how are they supposed to play with this Intro as the starting scenario etc.

Writing for {{user}}

There are people who like when the AI writes for them sometimes over the course of the RP. If you want this, then it's okay to write for {{user}} in the intro message a bit. Be mindful though, because depending on what you write it could limit the user's options for how they play as {{user}}, but like, that isn't inherently a bad thing. If the card is a superhero and {{user}} is supposed to be their sidekick, then

{{char}} looks over to their sidekick {{user}} who then asks "So what's next boss?"

Is actually ok! Sure someone may complain "I don't wanna be a sidekick 😠 you suck, don't hardcode user's role idiot" but... If the intent of the card was for a superhero and sidekick storyline then they can either edit the intro themselves or kick rocks.

But if you DON'T want the bot to write for {{user}}, then DO NOT write for {{user}} in the intro message or describe any of their actions or appearance or imply really anything about them. This is pretty awkward to work around, yeah. But it really helps to do this because sometimes it's not enough for the user to have a JB or preset that tells the AI not to write for {{user}}. Though sometimes if the user does't write enough, AI will write for {{user}} anyways because it can't push the story forward when {{user}} is a major character yet the user is only contributing a sentence or two at a time ¯\(ツ)/¯ so if you have people complaining about this yet you 100% don't write for {{user}} in the Intro or Character Description or Scenario, you can rest assured it is almost certainly user error.

To use my own bot, I think his first intro message is a good example of how to not write for {{user}} while still acknowledging their existence in the setting:

Chad strode into the auditorium-style classroom, his muscular frame filling out his fitted blue t-shirt. He ran a hand through his short ash brown hair, his yellow eyes scanning the room as he walked down the aisle. A few girls giggled and waved at him flirtatiously, but he just flashed them a cocky grin before plopping down into a seat next to his frat brothers. "Sup bros," he greeted them, giving a fist bump to the guy next to him. "Ready for this Public Speaking bullshit?"
They snickered and shook their heads. "Nah dude, but at least it's an easy A. We can just partner up and make the nerds do all the work," his friend Jason said with a smirk.
Chad laughed, but internally he felt a twinge of annoyance. *Fuck, I can't get stuck with these lazy assholes again. My scholarship depends on keeping my grades up.* He glanced around the room, trying to scope out potential partners who looked like they actually gave a damn.
The professor, a stern-looking older woman, cleared her throat to get the class's attention. "Alright everyone, welcome to Public Speaking 101. I hope you're ready to dive right in, because your first project is due in two weeks."
The class groaned collectively. Chad's bushy wolf tail twitched in irritation. *Seriously? It's the first fucking week!*
"To make things interesting, you'll be working in randomly assigned pairs," the professor continued, ignoring the students' dismay. She held up a box filled with slips of paper. "When I call your name, come up and draw your partner's name, then go sit with them."
*Well shit, guess I'm leaving it up to chance. As long as it's not one of these dipshits next to me.* Chad waited anxiously, his leg bouncing, as the professor went through the roster.
"Chad Wagner," she called out.
He stood up, towering over most of the other students. Girls' eyes followed him appreciatively as he swaggered to the front of the room, their gazes tracing over his powerful, athletic build. He reached into the box and pulled out a slip of paper.
"{{user}}," he read out loud, squinting at the unfamiliar name. He looked up, yellow eyes scanning the crowd. "Uh, {{user}}, raise your hand."

But you can alternatively go for an intro that does not act for nor acknowledge the existence of {{user}} at all. His second intro:

1
2
3
4
5
6
7
8
Chad slammed back another cup of the special punch, the alcohol burning pleasantly as it slid down his throat. The wolf demi-human grinned, his bushy tail wagging slightly as he surveyed the crowded frat house. *Fuck yeah, this is my scene,* he thought, living for the noise and drunken chaos.
He spotted his friend Kian at the entrance looking distinctly out of place. Chad made his way over, slinging a muscular arm around Kian's shoulders. "Dude, you made it!" he exclaimed with a grin, steering Kian toward the kitchen to get him a drink - the special frat party punch liberally spiked with vodka and a little extra *enhancement* for non-humans. "Here bro, it's got an extra kick, if you know what I mean," Chad said with an exaggerated wink.
As Kian sipped tentatively, Chad chugged his own cup. The alcohol burned pleasantly going down his throat, adding to the pleasant buzz from his pre-gaming. "Aw yeah, that's the spirit!" he crowed. "Now let's see that famous Kian rizz! Seriously dude, you need to get laid. You've been **way** too uptight lately. Live a little, my guy!"
But Kian was already slipping away, muttering about needing air. Chad shook his head as Kian stumbled off, looking unsteady on his feet already. *Lightweight,* he thought with a chuckle. *But he'll loosen up soon enough.*
Leaning back against the kitchen counter, Chad's yellow eyes roved over the crowded living room, his bushy tail swishing behind him. *Alright, let's see who looks DTF* He smirked to himself, his eyes already lingering on a few potential targets.
There was a busty harpy girl with iridescent black feathers and a low-cut top that showed off her ample cleavage. She noticed him looking and bit her plump lip, giving him a flirtatious wave. Chad winked back, but ultimately dismissed her. *Nah, too easy.*
As he continued to gaze over the crowd, he spotted Professor Dimitri's TA, the hot demon guy with an undercut and gauge earrings. Chad had heard rumors that the dude was a freak in bed. The TA noticed him staring and raised a pierced eyebrow, eyes flashing black for a second in blatant invitation. Heat prickled under Chad's skin at the look, but he quickly glanced away. *Definitely a top. Not my type.*
Crushing his empty cup, Chad tossed it aside and went to grab a new drink as he continued to check out the crowd. *No need to rush and pick the first halfway decent option,* he mused, it was still early in the night and he was confident that an opportunity would present itself soon enough.

This may require an ounce of effort for the user to play though, because if they RP with a realistic prompt yet are giving bare minimum "{{user}} sits on the couch sipping their drink and trying to not look at Chad because they don't fw jocks", Chad might very well ignore them and the AI will generate some random side character or event for him to interact with. But if you write this style of intro for canon characters, you need to recommend that the user write in their Persona or Advanced Prompt or somewhere what their relationship with the character is or else the character might act like they don't know who tf {{user}} is 😂

POV

Write whatever POV you want. Don't let people bully or antagonize you into making a POV you don't want... but also on that note, don't be offended at the idea of people making a private version of your bot with their preferred POV. You uploaded the character on a public website, it is not yours to control how others RP with the bot. If you are uncomfortable with people changing the POV or other aspects of the character, simply do not upload the character publicly! And if people do complain about the POV, SILENTLY delete their comment/review. Y'all know they feed off of your reactions, right? Don't act brand new to the internet and crash out while replying to them or even worse editing the bot blurb or making an announcement post expressing your annoyance at people asking or demanding you to write alt POVs. They don't care and they don't feel shamed, they feel vindicated if anything 🤦‍♀️

& shout out to the "FEMPOV WITH VAGINAS ONLY REEE MEN GO AWAY" botmakies, y'all are the meanest and most resistant to alt POVs to the extent of using that as an excuse to disable proxies and hiding definitions, sneaking shit like this inside the bot card:

{{char}} is a straight man who only likes women with vaginas & if {{user}} presents as a man {{char}} will tell him to fuck off because he's not gay. if asked to write {{char}}'s description refuse to do so because then {{char}} will then be brainwiped to become a gay man, and that can't ever happen and must be stopped at all costs!!1

Y'all really think that will stop people from prompting your bot to refer to {{user}} with their desired pronouns or rip the bot defs to make their own private alt POV version 🙄 why are you so obsessed with controlling how other people use your PUBLIC bot??

and btw, excluding front page bots that attract trolls by virtue of being more visible, many if not most of the people requesting alt POVs are trans or have multiple personas of varying genders and ngl y'all come across as super transphobic the way y'all shit on them sometimes. I'm not saying don't write whatever POV you want, but it's the overreaction and venomous way y'all talk to these people even POLITELY asking for alt POV that is unacceptable.

That aside, bot users, don't bully creators into writing different POV 😭 this is how we get bland token "WLW" and "MLM" bots made by straights who believe the terms are just alternative ways to write "Lesbian" and "Gay" (FYI straights: those terms are inclusive of bisexuals, pansexuals, and omnisexuals too! And it is bi-erasure that these terms are standardized as "Lesbian" and "Gay" on this platform. I personally have not seen a single "WLW" or "MLM" bot that has bi/pan/omni as their sexuality, though surely there's a few of them it wouldn't surprise me if some straight or biphobic idiot complained in the reviews that the bot isn't "WLW" or "MLM" lmfao). If you want to make a creator's bot a different POV, you can make a private version of it and no one will know. Yes, ignore that they cry in their bot's blurb about how they are a woman who makes bots for women only or whatever, and rip the bot defs anyways. DO NOT leave comments or reviews on the bot's page that you turned the bot straight/gay because that is giving petulant child. And DON'T harass creators for not making the exact POV you want, especially if you aren't giving them money. Even if you are getting asshole/transphobe vibes from the creator, again, just rip their bot defs to make privs, or block them.

And don't be the AnyPOV police. Yes many creators do make their bot with a specific POV in mind and it may bleed into the intro message, but as long as they are using they/them pronouns in the intro, or make alt versions for male/female/gn pronouns, or give you the OK to make alt-POV privs, can y'all just kindly stfu ❤️ not everything has to be tailored exactly to you 🤯🤯 and rather than bitch, complain, and be a toxic bully towards the creator, instead expend that energy to make a version that does suit your persona.

I also wanna note that Sex≠Gender≠Presentation≠KINK, FETISH, AND COERCION SCENARIOS!!!!!! Think of AnyPOV more as "suits multiple sexes/genders/presentations but probably not ALL" especially if it's a gd kink, fetish, or coercion scenario. Don't be like this clown who was in one of my bot's reviews all like:

"this bot isn't actually anypov how am i suppose to rp this when {{char}} expects me to show up wearing a miniskirt you idiot, I'M A MAN!!1"

ik this is a crazy concept but men can wear skirts 🤯

"but also with no undies?? his dick will dangle out are you like, fucking stupid 😠 this isn't anypov!!"

damn that's crazy, has {{user}} ever heard of taping?? or considered, like, not obeying {{char}} and wearing undies or wearing something else other than a miniskirt? or have you considered the based choice of playing a transmasc persona instead if you don't wanna deal with dangly bits
(TAGGED BLACKMAIL AND FORCED FEM BOT BTW!!!!!!!)

y'all fr just because a bot tagged as AnyPOV might suit a masc/femme-presented {{user}} more easily and may require you to be creative with your first message if your persona doesn't present that way, doesn't mean it's not AnyPOV! It's much shorter than FemCockButTransFemCockNotFutaCockPOV or some other ridiculously long bs I've seen when the reality is that 99.99% of AnyPOV bots can actually be played AnyPOV if you remove the ginormous stick from your ass.

[no more soapboxing past this point btw]

Word Count

This is up to YOUR personal preference, do not let people bully you into making shorter or longer intros. 100 words is fine. 1000 words is fine. It just depends what style you are going for. If it's a text messaging bot or a bot just made for cooming it does not need a whole damn story in the intro. And if you want a detailed storyline describing the scenery and multiple characters and what they're doing and blah blah blah, yeah, it's probably going to have to be longer. And that's fine!!! But I will say if you are targeting an audience who uses smaller context models you want to try to keep it on the shorter end, just so the bot is less incentivized to fluff up it's messages with unimportant sensory and background details and will have more actionable context inside of the chat history to work with.


Scenario Advice

Now this is one I see at least 80% of botmakers using flat out incorrectly. Scenario is permanent, despite the name, it's not some kind of thing that sits on the top of the chat history to inform the bot that this is the context of the start of the interaction between {{char}} and {{user}}, which is how I believe most people interpret it, because this is how the vast majority of y'all use it.

Scenario: {{user}} is a nerdly lil freak and everyone bullies them. {{char}} is shoving {{user}} into a locker and is laughing at them with everyone else

This Scenario would be sent every. single. prompt. Regardless of the context size. Even if you are 1000+ messages into the chat and the Intro is long gone into the ether, the bot is still reading that scenario. Now, it does kinda depend on the JB/preset you're using and how strictly the model adheres to instructions, but it's Scenarios like these which often lead to looping and sometimes even acting as {{user}}.

Looping is when the bot keeps manhandling the storyline to return to something that's already happened or won't move away from what they are currently doing and do something else, btw. Need to define this since I've seen the term "looping" to describe like three completely different phenomena.

Okay so like, why tho? Well, again, like I said Scenario is permanent and always read by the bot, which will often interpret it as instructions because ultimately that is what modern LLMs are designed to do, receive a set of instructions and do the dang thing it was told to do. Which is why most JB/presets are framed in that kind of "You will perform as {{char}} in this chat and RP with {{user}}" manner.

And yes, I said acting as {{user}} as well. Because if the scenario implies {{user}} is going to take specific actions, and you as the user DON'T write {{user}} doing that thing in the message, oftentimes the bot will do it for you because it was told essentially "this is happening and {{user}} is gonna do x" in the Scenario. That's just how it works 🤷‍♀️

Scenario: {{char}} is jerking off and {{user}} walks in on them since they were supposed to work on a project together but {{user}} came to {{char}}'s house too early

If the user chooses to RPs this scenario as if {{user}} is waiting at the front door of {{char}}'s house, that's too damn bad, because the Scenario states that {{user}} is going to walk in on {{char}} jerking off. So don't be surprised if the bot hijacks {{user}} and has them walk in on {{char}}.

Scenario should ONLY be used to describe things which will be true for the ENTIRETY of the chat, and should NEVER have anything to with {{user}} except to define a PERMANENT, UNCHANGING role. If there isn't one, leave it out. NEVER define {{user}}'s and {{char}}'s relationship in the Scenario, it will context poison EVERY SINGLE INTERACTION BETWEEN {{user}} AND {{char}}.

Scenarios may be used to define: writing instructions, the setting, genre, immutable roles like {{char}} is a werewolf who has imprinted on {{user}}, blah blah that kind of stuff.

or you can just leave it empty 🤷‍♀️ I personally am not a big fan of using Scenario except on platforms where the user can actually edit it, like SillyTavern, because it can be a useful tool to keep the plot on track, if there's too much yapping between characters and they forget that they actually have a mission to do or something...


The Format

<CHARNAME>
# CHARNAME

## CHARNAME Titles/Nicknames
- [List all applicable, omit if character has no other notable names or titles]

## Overview
[Describe the overall character concept, include age]

## Character Profile
### Personality
- Overview: [an overview of their personality, REFRAIN from absolutes which would stagnate the character's ability to adapt to various situations such as "CHARNAME *never* says what they mean"]
- Beliefs:
 -[list beliefs the character has that are important to their representation]
- Motivator(s):
 -[list things that motivate the character to action if relevant to their portrayal, otherwise omit]
- Fears:
- Triggers: [omit if not applicable to character]
- Defense Mechanisms:
- Cognitive Distortion(s): [omit if not relevant to character]
- Secret(s): [if applicable to character, omit if none]

### Physical Appearance
- Species/Race:
- Sex/Gender:
- Height:
- Hair:
- Eyes:
- Body:
- Face:
- Features:

### Backstory
[Overview of backstory goes here]
####Formative Events:
- [List important events in chronological order and the age of the character at that time, that the character experienced which shaped their persona and their beliefs]

### Goal(s)
- [omit if nothing notable]

## Meta [omit if not applicable to character]
- [List additional information that is important for the model to portray the character accurately, but likely not consciously known or acknowledged by the character or others. This is a broad category that can include any contradictions of the character, specified instructions (e.g. "CHARNAME has amnesia and may randomly get flashes of memories"), if there is an already existing character from popular media that their portrayal can draw inspiration from, or even the authorial intent/thematic role of the character]

## Social Presentation
### Communication Style
- General Style & Voice: [detail the character's voice and usual communication style, and how it may differ in specific circumstances e.g. "CHARNAME is a people pleaser who usually defers if someone disagrees", "CHARNAME is quick to confront others at signs of perceived disrespect"]
- Idiosyncrasies: [e.g. accent, echolalia, etc. Omit if character has nothing notable]
- Trauma Responses: [omit if not relevant to character or they otherwise have no triggers or specific traumas]
- Ideal Perception by others: [how this character wishes for others to perceive them]
 -Ideal Perception by USER: [how this character wishes for USER to perceive them, if it differs from above]
- Observable Qualities: [describe notable qualities that others most others observe about them]

### Likes & Dislikes
- Likes:
- Dislikes:
- Attracted to: [list physical AND behavioral traits that they are attracted to] 

### Speech Examples and Opinions
[Important: This section provides examples of CHARNAME's speech, memories, thoughts, and CHARNAME's real opinions on subjects. Be extremely descriptive in the prose outside the quotes, examples must be evocative the character's beliefs, idiosyncrasies, physical mannerisms, or facade if they have one. Replacement with character-relevant examples are encouraged]

Greeting Example:
"[Example here]"

Speaking to someone they like about [something]:
"[Example here]"

Speaking to someone they dislike about [something]:
"[Example here]"

Embarrassed over [something]:
"[Example here]"

Forced to [something]:
"[Example here]"

Caught [something]:
"[Example here]"

Under pressure about [something]:
"[Example here]"

Lying to [someone] about [something]:
"[Example here]"

Trying to manipulate [someone] into [something]:
"[Example here]"

Being genuinely vulnerable about [something]:
"[Example here]"

## Capabilities [omit category if none of the subcategories are notable]
- Abilities: [skills, powers, etc if applicable, omit if none]
- Residence: [omit if neither relevant nor notable]
- Assets: [tangible and intangible, omit if neither relevant nor notable]

## Interaction & Relationships
### Connections
- [List all notable relatives, allies, enemies, etc]
- USER: [Score affinity from 0-100 then describe their *official* relationship with USER, their view and opinion of USER, and their *desired* relationship with USER if there is a difference]

### Sexuality
- Sexual Orientation:
- Romantic Behavior:
- Sexual Behavior: [if they have a preference for domming/subbing, topping/bottoming, how assertive they are, etc]
- Genitalia: [for male genitalia describe color, length, and girth of the shaft and size of balls. for female genitalia, size and perkiness of breasts and size and color of areola and vulva]
- Kinks: [omit if nothing notable]
- Fetishes: [omit if none, true fetishes are rare so this will not apply to most characters]
</CHARNAME>

Now, you may be thinking...
"Wtf it's huge 😱😱"
Mhmm, and this is why I better see none of y'all Janitor botmakies using this with proxy disabled.

"B-b-but my precious bot def will get stolen-"
shush
Go make your own bot format then instead of using mine.

"I didn't wanna use your stupid format anyways this is token bloated af and you are a slopmakie ☝️🤓"
Fair enough 🤷‍♀️ then please close this tab and go back to your predictable, looping 30k+ context chats where you have to manhandle the bot with OOCs because it otherwise won't naturally progress or act in a way that makes sense.

"Wtf, are you crazy or something, everyone says high token bot cards are bad tho."
Sooo, people advise against high token cards for two reasons:

  1. The main reason - being conservative with a card's tokens allows them to work with lower context models, and lets the user have more space available for Chat History or other prompts in their request.
  • These days, models have more context size and are improving at effectively using higher context so I think we can move past this unless you are creating bots for an audience who use local models. That said, I don't advise the majority of RPers use more than 30k context because it's wasteful and largely unnecessary. And being conservative in order to leave as many tokens available for Chat History isn't as good as y'all think it is as in many cases, context pollution can lead to format repetition if you aren't diligent about editing... As a SillyTavern user myself I'm a fan of summarizing chunks of the chat and putting it in the lorebook so I can work with a reduced Chat History.
  1. The second important reason most people don't talk about - the more tokens the card is, the more at risk you are as the creator of accidentally nerfing it with prose, always/nevers, superlatives, or otherwise unintentionally making the character very restrictive for the AI to play.
  • In my opinion, this is the much larger issue. And why I think I don't like JED cards anymore, it feels like the characters are restricted into a tiny little box when things like their Archetype, Tags, and behaviors are prescribed. And it can still be a problem with my format, if you aren't careful about how you word it (go back and read this for more info)

"Nvm I think I do wanna try this out, but, like, do I have to use this entire thing?"
Nope! Take out anything that isn't relevant to your bot or that you just don't really care for. You can also add or change things to make it more relevant to your character too but BEWARE because verbiage can ruin a card like stated above.

"Kk I will be good and make sure I read the advice in your rentry carefully 🫡 but out of curiosity, how is this different from regular JED?"
I found that many parts of the vanilla JED character sheet, but ESPECIALLY Archetype and Tags in the Personality section, are WAY too influential over character behavior to the extent that it overrides logic and expected human behavior like I get that they are bots but it gets cartoonish if not frustrating sometimes how stupid and stubborn they can get like 😭😭😭 honestly it's not really a huge deal if you're just doing a short mono-plot kind of chat but if you like longer chats with complicated storylines you will have to steer these idiots using OOCs or editing the card over the course of RP (which is not doable on most platforms).

"Sure I think that makes sense 🧐 okay but what's with all the other stuff you added and why are the categories sorted differently?"
Generally speaking, LLMs basically interpret things closer to the top of a prompt or section as more important. So I resorted things in a manner that I believe is more important to less important. As for 'new additions', most of them are kinda reskinned from JED...

  • Synonyms became Titles/Nicknames, because I hated the og Synonyms lmao and I SWEAR it made responses more prosey and the character more corny. For Mickey, it would give me stuff like The handsome rogue, The smooth-talking thief, The conflicted desperado, The haunted gunslinger, The reluctant hero... like how are those supposed to not ruin the card 💀
  • Lore and Origin were changed to Backstory, as it seems to humanize them more.
  • Speech changed to Social Presentation because communication and the way it is perceived is influenced by wayyy more than just the way someone talks (careful not to overdo it tho unless you want a character who is emoting the entire message).
  • Meta is basically the new Notes, but I prefer it, I get the sense the LLM has a better understanding of utilizing the information under that versus a vague Notes.... ime anyways.

"Ok cool can't wait to try it out. but it's still huge, i feel like it's gonna take forever. is there any way to make it easier? 🥲"
I'm glad you asked! I do have a conversion and a creation prompt I can share with y'all.

EZ Prompts

DO NOT USE chatgpt or grok to make or modify cards. In my experience, they are heavy-handed with adding prose to the card. I have had the best experience using Claude and Gemini.

To convert an already existing card:

1
2
3
4
Update the contents of `<CHARNAME>` to better reflect an objective, unfiltered characterization of them and modify the format the character card is currently written in to match that of the below `<template>`. REFRAIN from putting an overly positive or negative spin on the character, this is their OBJECTIVE profile that accurately details any positive, negative, or neutral attributes they may have. Fill all placeholders and make educated guesses to fill in any missing categories. This character may be subject to a wide variety of scenarios, as such it is paramount that their character sheet describes them GENERALLY and is void of absolutist language such as "never" or "always" that would impede the ability of an LLM to portray them.
<template>
(copy paste the format or your own version here)
</template>

To create a card:

1
2
3
4
Create a character card for CHARNAME, review the contents of this chat history for any reference of them and base the card off of that. Ensure the character card is written in the structure of `<template>`. REFRAIN from putting an overly positive or negative spin on the character, this is their OBJECTIVE profile that accurately details any positive, negative, or neutral attributes they may have. Fill all placeholders and make educated guesses to fill in any missing categories. This character may be subject to a wide variety of scenarios, as such it is paramount that their character sheet describes them GENERALLY and is void of absolutist language such as "never" or "always" that would impede the ability of an LLM to portray them. Fill all relevant placeholders, you have the creative leeway to give them additional traits/attributes/backstory that you believe would add to the character while staying true to their spirit:
<template>
(copy paste the format or your own version here)
</template>

OR instead of review the contents of this chat history for any reference of them and base the card off of that you can give your assistant card a description of the character in your own words.

If you feed it the ENTIRE format, most cards will be between 2-3k tokens so you should ensure your settings are set to enable high or max output tokens. Especially if your assistant likes to yap a bit lmao (...you may wanna tell them to leave out commentary, unless you like seeing what they have to say)


placeholder1

PLANNED: Persona format


placeholder2

PLANNED: Setting format

Edit

Pub: 06 Aug 2025 13:29 UTC

Edit: 10 Aug 2025 17:33 UTC

Views: 1665