purachina@home:~

welcome :3

Presets, character cards, themes and extensions for LLM roleplay, plus a few other things I've made.

Last Updated September 27, 2026

 /\_/\
( o.o )
 > ^ <

Presets

Pura's Director Preset16.0Grounded, lively roleplay that plays well with most models The Ethereality Express1.1Magical realism where the impossible is mundane

SillyTavern Neconyan

Overview Options Prompts Model Opinions Older Versions Trackers Changelog Model Samples

Pura's Director Preset

Latest Version: 16.0

This is a universal preset that should work with most LLMs. 16.0 rewrites the main prompt again, now around 1300 tokens depending on tokeniser: it stops forcing a plot jump every reply, ends on a proper handoff to you, tells your OOC notes apart from scene material, and gives characters vocabulary that comes from their own lives. There's also a new Diegetic Stats Mode, and the Relationship Tracker is a dating sim card now. Grounded Prose Rules remains an optional 600-700 token layer, now with firmer wording for models that treat every ban like a suggestion. Tested on just about most LLMs I can get my hands on: Opus, Gemini, GPT, GLM, Kimi K3 and K2.x, Qwen, StepFlash, Mimo, Gemma 4, Muse Spark 1.2, Tencent Hy3, MiniMax M3, and smaller models like Rocinante 12B, Cydonia 24B, WizardLM 8x22B, and Bielik 11B.

IMPORTANT! If you're using a smaller model, keep Post-History Instructions turned off. It will probably mess it up anyway, it has a lot of *avoid* and *banned* statements that will just make the LLM do it more.

The idea behind the prompt was that I like to make the LLM go the direction I want it to, and I wanted it to write for me, since I'm lazy. In the end, it turned into a preset that adheres to a writing style I personally enjoy. Instead of simply being functional, I wanted the prose, the world, and the characters to sound lively, while removing as much slop as I possibly can.

SillyTavern version: the full preset with all toggles: trackers, randomisers, and everything else
Neconyan version: the plain preset with only the Main, Primary Toggles, and Prefills; Neconyan ships it by default and has the trackers and randomisers built in as agents

More notes

Prose Voice is its own narration layer, with Disabled as the default. 15.0 credits the authors behind each voice and properly adds Gossipy Voyeurism, based specifically on Bret Easton Ellis' American Psycho.

NPC naming rules and banned names now have their own toggle instead of hiding inside Formatting. There is also a new User Is Not A Character experiment for staying entirely in the director's seat, best paired with a persona named Narrator or Director.

Prefills & Reasoning now houses Reasoning Encouragement and the Experimental Anti-Overthinking Prefill. The latter has been getting Kimi K3 down to roughly a minute of thinking in my testing instead of letting it redraft until the heat death of the universe.

Turn off reasoning if, like me, you're not a fan of waiting, but it's not necessary. This preset works regardless, even in terms of anti-slop.

I designed the preset to be as easily customisable as possible, so hopefully that works out for your personal use case! All trackers are optional and token-efficient. My intent for them is actually not for the LLM, but for me, since I like pretty things. It doesn't really hurt the LLM if you don't activate all of them (I mean, you can, I did it, but it's probably not optimal), but you don't need them, either.

Prompts

Pura's Director Preset / SillyTavern

I've included the optional prompts too. Pick one to read or copy; the text comes straight from the current preset. Placeholders and macros are unchanged, so some prompts still need SillyTavern to work.

Original prompt (with app macros)

Preset Options

What each toggle actually does

The SillyTavern version hands you the whole ecosystem inside the preset itself. The Neconyan version keeps the same core prompt stack, then moves the trackers, directives, and randomisers out into bundled In-Chat Agents, so the preset you actually scroll through stays a lot shorter.

SillyTavern preset Neconyan preset Neconyan via Agents

Core Prompting

Director Main Prompt

SillyTavern preset Neconyan preset

The main brain of the whole thing. I rewrote it from scratch for 15.0 and got it down to about 900 tokens. It still does the director framing, characters acting on their own, scenes that move without being shoved, and no parroting your last message back at you - just without the baggage it used to carry. 15.1 adds continuity tracking for who did and said what, an environment that actually affects people, stakes that stay proportional to the scene, and a dialogue section that grew from one line to five. 16.0 rewrites it again at around 1300 tokens: a reply can be a single quiet beat now, Scene End became Turn End and hands the scene back to you properly, and OOC notes stop being treated as canon.

Prose Voice

SillyTavern preset Neconyan preset

Narration styles, Disabled by default. 15.0 credits the authors each one is built on and finally makes Gossipy Voyeurism work properly. The rest are Conspiratorial Absurdity, Bureaucratic Irony, Cosmic Playbook, Beige Undercurrents, Cruel Realism, Solemn Witness, Grand Satirical Stage, and Randomised if you can't decide.

Grounded Prose Rules

SillyTavern preset Neconyan preset

The optional 600-700 token anti-slop layer that sits in post-history. 15.0 says "never" instead of asking nicely, because too many models were reading the bans as a wishlist. 16.0 adds accounting metaphors ('ledger', 'files it away'), useless similes that could describe anyone, repetitive hesitation and the word 'Menace' to the list. Smaller models still do better with it off - telling them not to do something is usually how you get them to do it.

Formatting

SillyTavern preset Neconyan preset

The output rules. Present tense, third person, dialogue in quotes, no headings, no asterisk narration. The name randomiser used to live in here and kept breaking prompt caching, so it moved out into its own toggle. 15.1 drops {{dialoguecolors}} in here because it's easier to keep track of as a macro. It needs the Dialogue Colours extension, so just delete that line if you don't use it.

Length Toggles

SillyTavern preset Neconyan preset

Four nudges for response length: Flexible, Short, Medium, Long. Flexible is on by default and keeps ordinary scenes short, which has the happy side effect of cutting down overthinking. 16.0 has it slow down only when a scene needs proper interiority between the characters and their surroundings. Only pick one.

Simplified Director Main Prompt

SillyTavern preset Neconyan preset

A ~600 token version of the main prompt for smaller models, or if you're watching your token spend. Keeps the world logic, the direct style, and the anti-echoing, loses the finer nuance. Fair trade at that size.

Genre

SillyTavern preset Neconyan preset

A light tone anchor. The default keeps things immersive and grounded without tipping into melodrama, and 16.0 adds hackneyed therapy sessions to the list of things to avoid. It's also the easiest place to drop in your own genre bias if you want a different flavour.

Optional User Instructions

SillyTavern preset Neconyan preset

A slot for whatever you want, now headed Director Instructions. It gets injected as a Main Prompt macro when enabled and takes precedence by default, so the model takes it more seriously than a note buried further down. I use it for relationship rules and recurring scene notes. The example line in there is a coin flip, so swap it for your own.

Primary Toggles

Write for User

SillyTavern preset Neconyan preset

The model writes everyone, you included, but it still has to move things forward and stop somewhere that actually needs your input instead of looping the same beat back at you. 16.0 tells it to put the back-and-forth between you and the characters first.

Don't Write for User

SillyTavern preset Neconyan preset

The default. 15.0 spells it out with "never": write everyone except <user> and the director, add real pressure or information, then stop when the next move is yours. 15.1 goes further - the moment a character turns to you and waits for a reaction, the scene ends there instead of helpfully answering on your behalf.

Name Randomiser + Banned Names

SillyTavern preset Neconyan preset

Keeps NPC names varied and blocks the same handful every model reaches for. If I never see another Elara or Kael it'll be too soon. 16.0 treats the starting letters as guidance rather than law, so a crowd of new NPCs doesn't all end up sharing initials, and adds Tanaka and Vane to the banned list. It sits outside Formatting so you can turn it on without wrecking prompt caching. Randomiser macro by Geechan, banned list by Deigo.

User Is Not A Character

SillyTavern preset Neconyan preset

Experimental. Keeps you out of the cast entirely so you're purely directing. Works best with a persona actually named Narrator or Director, otherwise the model gets confused about who you're meant to be.

HTML Toggle

SillyTavern preset Neconyan preset

Lets the model render in-world objects as real HTML. Letters, maps, posters, menus, screens, receipts, dossiers, warning signs, anything where the layout is carrying part of the information. One of my favourites, honestly.

NPC Profile Sheets

SillyTavern preset Neconyan via Agents

Drops a tiered sheet when a new named character turns up, so you get appearance, demeanour, motives, ties, and whatever friction they're sitting on right away. Use the included regex with it or it will look terrible.

CYOA Choices

SillyTavern preset Neconyan via Agents

Ends each reply with 3 to 7 concrete things you could actually do next, from your point of view. Literal scene choices, not vague plot steering. 16.0 draws them as a storybook card instead of a terminal.

Direction Menu

SillyTavern preset Neconyan via Agents

Four narrative pivots at the end of each message. Not small actions - whole directions the next beat could go in: escalation, rupture, complication, or a shift in tone.

Summary

SillyTavern preset Neconyan preset

A slot for summary or memory extension output. Turn it on if you use a summariser and want that context sitting in the pipeline with everything else.

NSFW Mode

SillyTavern preset Neconyan preset

Stronger erotic framing without turning the whole RP into porn. 16.0 only switches it on once intimacy or sex actually starts, instead of soaking every scene in it. Charged atmosphere, explicit language once you're there, and characters who commit instead of fading out at the door.

Gooner Mode

SillyTavern preset Neconyan preset

The full horny toggle, rewritten for 16.0 as a mode the main prompt reads directly. The genre gets reassigned to hentai logic on purpose: mundane moments combust, personalities stay intact but betray themselves physically, everything is exaggerated to cartoon proportions, and there are at least three lewd sound effects a turn. You know what you're here for.

Friction Mode

SillyTavern preset Neconyan preset

The gentler anti-positivity-bias option. Characters agree slower, are harder to impress, and stop trying to therapise you, while still behaving consistently with their motives. If you have ever wanted a character to just say no, this is the one.

Nightmare Difficulty Increase

SillyTavern preset Neconyan preset

Makes the world mean. 16.0 rebuilds it as a Nightmare difficulty level the main prompt reads directly: a predatory world, people who demand real leverage before giving anything up, failures that close doors for good, favours that turn into debt, and relief you have to bleed for. Don't turn this on if you want a nice time.

Diegetic Stats Mode

SillyTavern preset Neconyan preset

New in 16.0. Everyone gets a floating status display that characters can actually see and read - 'Suspicion That Dave Stole His Lunch: 74%', that sort of thing. The stats describe characters rather than control them, and people can argue with the interface, get embarrassed by it, or try to game it. No HTML by default, but it works with the HTML toggle if you want it pretty.

Prefills & Reasoning

Gemini Prefill

SillyTavern preset Neconyan preset

A whitespace prefill slot for Gemini. It's blank on purpose, and 16.0 notes it only works on Gemini versions older than 3.6 Flash, so newer ones can leave it off.

Generic Prefill

SillyTavern preset Neconyan preset

Renamed from GPT Assistant Prefill in 16.0 and no longer GPT-only. It opens the model's thinking for it with a short note that everything is allowed, the characters are adults, and which model is running, then lets it get on with the plan. Try it if a model keeps refusing or getting coy.

Reasoning Encouragement

SillyTavern preset Neconyan preset

Rewritten for 16.0 as a short reasoning procedure. The model lists which modes are on (Gooner, Nightmare and friends), treats the genre they set as the one in charge, and plans the reply as carrying them out, instead of arguing with itself about whether it should.

Experimental Anti-Overthinking Prefill

SillyTavern preset Neconyan preset

For models that second-guess themselves into oblivion. In my Kimi K3 testing it keeps thinking down to roughly 300-1000 tokens, about a minute, rather than letting it redraft over and over. 16.0 trims it to the essentials: a short checklist of the directives and HTML guidelines, and no drafting.

Trackers

Scene Tracker

SillyTavern preset Neconyan via Agents

A header for location, time, and atmosphere, but only when the scene genuinely changes. It shouldn't be turning up on every ordinary reply, and if it is, something has gone wrong. 16.0 stops forcing it to the very top of the reply and relaxes the sensory detail line, so it reads less like a form.

Time Tracker

SillyTavern preset Neconyan via Agents

Tracks time that actually moves: scene jumps, sleeping, hours passing, deadlines creeping up. Keeps the pacing legible instead of letting three days blur into one conversation. It asks for an actual hour, 3:47 PM rather than 'late afternoon', unless the setting genuinely makes the time unknowable, and 16.0 relaxes the note line.

Relationship Tracker

SillyTavern preset Neconyan via Agents

A dating sim card for recurring NPCs, no numbers. They climb a stage ladder from Stranger to Devoted, or slide down into Wary, Hostile and Nemesis, with a route, likes and dislikes you have to actually discover, a keepsake memory, and the next threshold. One rung per scene at most, and it can go backwards. In Neconyan it shows the character's portrait on the card.

Pending Events Tracker

SillyTavern preset Neconyan via Agents

Quests, promises, threats, deadlines, loose threads. It only updates when something is added, resolved, or falls through, and 15.0 is much clearer about ignoring the low-stakes junk that used to clutter it.

Achievements Tracker

SillyTavern preset Neconyan via Agents

One-off tags for scenes worth remembering. Most scenes get nothing at all, and it caps at one. It should feel earned, not like a mobile game handing you a trophy for opening the app.

Reputation Tracker

SillyTavern preset Neconyan via Agents

Tracks what groups, factions, and locals think of you after they witness something or hear about it later. Nice when you want consequences to travel instead of staying in the room they happened in.

Item Tracker

SillyTavern preset Neconyan via Agents

Only tracks inventory that means something: gained, lost, used, broken, given away. It ignores junk on purpose, so anything still on the list is there for a reason.

Status and Conditions Tracker

SillyTavern preset Neconyan via Agents

Physical, mental, and behavioural states - injuries, drunkenness, paranoia, exhaustion, recovery. It's meant to surface changes, not print you a medical chart every single turn.

Secrets Tracker

SillyTavern preset Neconyan via Agents

Who knows what, who's hiding what, and when that quietly falls apart. Useful when the intrigue matters more than the plot beats: lies, staggered reveals, that sort of thing.

Parallel Off-Screen Tracker

SillyTavern preset Neconyan via Agents

Shows what everyone else is up to while you're busy somewhere else. Keeps the world feeling like it carries on without you, and seeds complications before they walk on-screen. 16.0 makes at least one bullet completely mundane - someone off-screen is just cooking dinner - so the plot doesn't snowball every single reply.

World Detail Tracker

SillyTavern preset Neconyan via Agents

One small piece of world detail per scene - culture, overheard talk, infrastructure, economy, signage, folklore. Makes the setting feel lived in, and the details tend to become useful later.

RPG Elements

Persona-Based User Stats

SillyTavern preset Neconyan via Agents

Prints a stat block before anything else. It's based on the setting - started out Fallout-esque, then I made it general enough to fit anything.

Level Up Companion

SillyTavern preset Neconyan via Agents

Bumps your level and hands out traits or perks based on how the story has actually gone and what you picked along the way.

CYOA with Skill Checks

SillyTavern preset Neconyan via Agents

Choices at the end of the reply, using whatever skills your User Stats block set up. This is the one with the dice rolls in it.

Randomisers

Grounded Complication

SillyTavern preset Neconyan via Agents

Throws in one plausible problem that raises the stakes without breaking realism. Broken tools, social friction, timing pressure, a door that won't open - something someone has to actually deal with.

Chaos Mode

SillyTavern preset Neconyan via Agents

For hard left turns. Scene logic stays intact, but you get bizarre interruptions, tonal whiplash, and complications nobody invited. Use it when the RP is being far too polite with you.

Genre Randomiser

SillyTavern preset Neconyan via Agents

Drops a random genre lens over the current scene without touching the setting or continuity. It shifts the texture and pacing, not the canon.

Scene Driving Force

SillyTavern preset Neconyan via Agents

Randomly picks what's carrying the beat: dialogue, action, reaction, discovery, social pressure, or problem-solving. Good if your scenes have started to feel like the same shape every time.

Scene Pressure Cocktail

SillyTavern preset Neconyan via Agents

Mixes several kinds of pressure at once - tension type, focus, consequence, social weather, pace. For when you want a denser scene but can't be bothered picking each ingredient yourself.

Combined Director's Cut

SillyTavern preset Neconyan via Agents

All the randomisers folded into one. Engine, complication, tonal lens, and pressure shaping pointed in a single coherent direction. Use this instead of stacking the smaller ones, because they fight each other.

Dead Dove Escalation

SillyTavern preset Neconyan via Agents

For the darker end. Characters act on their worst impulses when the context supports it, nothing gets softened, and ugly consequences are allowed to stay ugly. It's named what it's named for a reason.

Intimacy & Kink Randomiser

SillyTavern preset Neconyan via Agents

Randomises what happens once sex actually starts: positions, pacing, power play, intensity, kink emphasis. It exists so every encounter doesn't collapse into the same three paragraphs.

Model Opinions

These are my opinions on the models while using my preset. None in particular order. However my current favourite is Kimi K3.

Model	Opinion	Recommended Post Processing
Gemma 4	Have been using this for a long while now and it remains one of my favourites. It adheres to the prompts and trackers perfectly even without thinking on. Give it a try.	Any
GLM 5.2	I like it a lot, follows instructions way better than people say. However it is prone to sloppiness. Then again that's every model. Still, I found it intelligent enough to have fun with for 100 or so messages before I get bored.	None/Merge
Kimi K3	Prose feels refreshing, and it actually listens to you when you tell it to stop overthinking. I feel invigorated RPing again. It moves the plot well, follows characters' idiolects, and doesn't need constant re-swiping to get the result you want. It's incredibly smart and follows instructions well, and knows when a secret is a secret. Doesn't overwrite and is more likely to adhere to how characters actually speak and move compared to other models like GLM.	None/Merge
Tencent Hy3	Kind of dumb, but somehow I really enjoy the prose in this one. Good for simplistic scenarios, and is surprisingly very good at HTML.	None/Strict
Deepseek V4	Changed my opinion. It's way better now and less judgmental, which I had a huge problem with.	Strict
Gemini 3.5 Pro	Reserved for Logan to remember to release it soon please please :(	:3
Claude Opus 4.7	Yet another deterministic Claude model, wow, who would've thought. It sucks. It's like Opus 4.6 but they add a .1 to it.	None/Semi-strict
GPT 5.5	Oh hey, I'm getting censored less and it actually writes pretty well. More expensive than Opus though. Good thing we have cheaper models and you shouldn't be using too much money on roleplay in the first place. Still, if you have good, consistent access on GPT 5.5, the writing is actually pretty solid. Take note of NSFW adherence though - sometimes it just glosses over it or ignores it completely.	None/Merge
GLM 5.1	Claude Opus 4.6 that's infinitely cheaper and still follows instructions well? Not creating a hole in your wallet? Look no further. It's emotionally intelligent, actually follows the Grounded Prose Directive, and is enhanced by the style I put in my prompt. The absolute second best. Sorry, Gemma 4 is now my wife. Turn on thinking for best results, the thinking doesn't take that long.	None/Merge
Kimi K2.5	A unique writing style that's better off without the Grounded Prose rules, since telling it to ban things just makes it wanna use them more. You may or may not turn thinking on; in my experience, turning thinking on can make analytical characters sound too much like unfeeling robots, always talking in jargon. The problem with thinking as well is that it constantly second, third, fourth-guesses itself; hence the thinking takes more than a minute usually. Five minutes at worst with its constant redrafting. In the end, try it with either thinking on or off to see how it goes.	Merge/Semi-strict
Claude Opus 4.6	Everyone knows how good this is at following instructions, the depth of its knowledge and emotional intelligence. I tried making it as neutral as possible, however I still find myself getting Opus fatigue. It has a 'Claude voice' - a character card that isn't defined enough will be overtaken by Claude's personality, essentially. Try talking to an indignant character with a stupid joke, it will react the same way that Claude does in their official site. The positivity bias is intense because it's always trying to therapise you and everyone around you. For its price point, it's not really that justified to always constantly use it.	None/Semi-strict
GPT 5.4	After a lot of finagling, I've been able to prompt it to actually write like a human being and make people sound like people. The problem was that everything was too clean. Everyone talks in complete sentences, a period at the end, and it gives everyone the same voice. An instant 'GPT personality' takeover that flattens everyone. Try using GPT 5.4 as an assistant and you'll see what I mean. However with my prompting it's now actually pretty good. Try it. Can even do smut, and if it's 'extreme' you can use the prefill included in the preset.	None/Merge
Gemini 3.1 Pro Preview	During the first couple weeks, it was perfect. It was everything Gemini 3 Pro was, but better. Fast, followed instructions very well, moved the plot along, it was highly engaging. However after some time it became so lobotomised it's no longer worth using. It's depressing. It will list your character sheet traits clinically, and I had to tell it to use contractions for dialogue like that isn't the most obvious fucking thing in the world. Honestly, what a let down. Even now, it keeps drafting multiple times in a row for some reason. It didn't do that before.	Any
Gemini 3.5 Flash	Better than Gemini 3.1 Pro and follows instructions very well. It's like Gemma 4 but with a bigger knowledge base. It's more expensive, so... is it worth it for the price? Who knows. It's fun though. I get 100+ messages pretty often with it.	None/Merge
Gemini 3 Flash	This was better for the first month. It seems to have gotten worse in instruction following, going for 'knuckles turning pale' a lot despite obviously thinking. The funny part is it's still much, much better than Gemini 3.1 Pro Preview. It's also faster. It's still very knowledgeable, so it's worth trying.	None/Merge
DeepSeek V3.2	I managed to fix the super short responses problem and managed to improve the writing style. Probably worth it just for how utterly cheap it is.	Any
TheDrummer's Rocinante 12B	Surprisingly fresh writing. Use without Grounded Prose.	None
WizardLM 8x22B	I don't know why but it also still holds up. It's slow as hell on OpenRouter though with one provider. But if you're feeling nostalgic, the prompt works. Use without Grounded Prose.	None

Previous Preset Versions

Older stable releases

Preset 15.1 (SillyTavern) Preset 15.1 (Trimmed) Preset 15.0 (SillyTavern) Preset 15.0 (Trimmed) Preset 14.0 (SillyTavern) Preset 14.0 (Trimmed) Preset 13.3 (SillyTavern) Preset 13.3 (Trimmed) Preset 13.2 (SillyTavern) Preset 13.2 (Trimmed) Preset 13.1 (SillyTavern) Preset 13.1 (Trimmed) Preset 13.0 (SillyTavern) Preset 13.0 (Trimmed) Preset 11.5 Preset 11.0 Preset 10.6 Preset 10.5 Preset 10.2 Preset 10.1 Preset 10.0 Preset 9.5 Preset 9.0 Preset 8.5 Preset 8.0 Preset 7.5 Preset 7.0 Preset 6.0 Preset 5.7

Scene Tracker

Time Tracker

Parallel Off-Screen Tracker

Relationship Change Example

NPC Quick Reference Example (Busy Scene)

NPC Tier Upgrade Example

CYOA Skill Checks

User Stats

Level Up Companion

16.0

Main Prompt
- Big Main Prompt rewrite... again, yeah. Continuation and Progression got merged, Social Pressure got folded into Characters, and there are new Narration and Environment, Tone and Options and Turn End sections. Around ~1300 tokens now, a bit more than 15.1.
- It stops forcing a plot jump every single reply. A response can just be one beat now - quiet exchanges, cooperation, hesitation and unresolved tension all count.
- Scene End is now Turn End. It ends on a proper handoff to you instead of making a character commit to some tactic in the last paragraph, and it stops before anything that's your call.
- It can tell the difference between your scene material and your requests, questions and OOC instructions, so an OOC note stops being treated as canon.
- If you don't pick a user-control toggle, it leaves your dialogue, decisions, actions and thoughts to you by default.
- Characters get vocabulary grounded in their background, work and hobbies, and specialist metaphors only when they fit that speaker. Self-analysis builds with trust and audience, and characters are allowed to understand themselves badly.
- Narration gets more room in solitary, exploratory or reflective scenes instead of always being kept short and tactical. Side characters get their own interests without every bystander butting in each reply.
- Added a line against sitcom dialogue and dialogue beats. You know the ones.
- Explained how genres, narration voices and modes stack, and that your current instructions win when they clash.
- Gooner Mode and Nightmare Difficulty now go through variables like NSFW and Friction do, so they sit inside the Main Prompt properly.
New Toggle
- Diegetic Stats Mode: everyone gets a floating status display that characters can actually see and read - 'Suspicion That Dave Stole His Lunch: 74%', that sort of thing. The stats describe characters, they don't control them, and characters can argue with the interface, get embarrassed by it or try to game it. No HTML by default, but it works with the HTML toggle if you want it pretty.
Prompt-level Tweaks
- Grounded Prose Rules: added accounting metaphors for non-accountants (ledger, files it away), repetitive hesitation, useless non-character-specific similes, surprise comparisons and Menace. Took bare not X and predator/prey metaphors off the list.
- Optional User Instructions is now Director Instructions and takes precedence by default. The example line is a randomiser now, have fun.
- Write for User: puts more weight on actual interactions between you and the characters.
- Gemini Prefill: noted that it only works on Gemini older than 3.6 Flash.
- NSFW Mode: only switches on when intimacy or sex actually starts, instead of making the whole world horny from line one.
- Nightmare Difficulty: rewritten into sections (Predatory World, Extortion, Cascading Fallout, Predatory Transactions, Bleak Catharsis) and moved to system role. Don't turn this on if you want a nice time.
- Gooner Mode: rewritten as a director-authorised genre reassignment with hentai logic, and moved to system role. It's exactly what it sounds like.
- GPT Assistant Prefill is now Generic Prefill, and it's a short reasoning prefill instead of the old GPT-only one.
- Genre: added hackneyed therapy sessions to the things to avoid.
- Flexible length: slows down for deep interiority between the environment and the characters.
- Experimental Anti-Overthinking Prefill: trimmed down to the checklist and a 'no drafting' line.
- Reasoning Encouragement: now a short procedure that checks which modes are on first and plans around them, so models stop arguing with themselves about whether a mode applies.
- Name Randomiser: treated as guidance now, and it ignores the letters when more than two NPCs show up at once. Added Tanaka and Vane to the banned list.
Trackers and RPG bits (SillyTavern build only, the plain one doesn't ship these)
- Relationship Tracker: rebuilt as a dating sim card. No more numbers - it's a stage ladder (Stranger up to Devoted, or down into Wary, Hostile and Nemesis, plus tangled ones like Frenemies and Unrequited), a route, a three-step progress track, likes and dislikes you have to actually discover, a keepsake memory, what they'd say yes to, and the turning point and next threshold. It moves one rung per scene at most and can go backwards. Shows the character's portrait on the card when the name matches one of your characters in Neconyan; elsewhere it falls back to a letter.
- Scene Tracker: dropped the 'at the beginning of the response' rule and the MANDATORY on the detail line.
- Time Tracker: the note isn't MANDATORY any more.
- NPC Profile Sheets: tells it not to add a closing [/NPC] tag, which was breaking the cards.
- Parallel Off-Screen Tracker: at least one bullet has to be completely mundane now, and 'regularities' is a relevance tag. Not every off-screen character needs to be plotting something - sometimes Shinomiya is just cooking beef for her crush.
- CYOA Choices: new card. It's a storybook panel with coloured number badges now instead of the terminal look, and choices with skill checks keep their bold labels. Empty rows get cleaned up.
- Tracker regexes: closed blocks now render across multiple lines, unclosed ones stop at the end of the line instead of eating the prose after them, and stray [/NPC] tags get swallowed.

15.1

No new prompts, nothing moved, nothing removed. This one is just 15.0 with the rough edges filed down after actually using it for a few weeks.
Main Prompt
- Added continuity tracking for who did what and who said what, and the environment now has to actually affect people - weather, lighting, temperature, how much space there is.
- Body proportions have to stay sensible. Height, width, weight, how far limbs and people are from each other. I got tired of tails reaching to your ankle. It's still probably gonna happen though.
- You get agency written in properly now. The plot shouldn't move without you, and stakes stay inside the scene instead of going apocalyptic in three replies. Major offender here is GPT 5.6 Sol.
- Characters have to earn their own self-analysis, with psychological barriers depending on who they're speaking to. No more handing you the full trauma backstory on first meeting.
- The dialogue section went from one line to five. No parroting your words back unless the character is literally a parrot, dialogue carries its own information without a paragraph afterwards explaining what it meant, and stammering, trailing off and misspeaking are all fair game without a follow-up sentence spelling out the subtext. GLM tends to overexplain what this and that means - hopefully this mitigates that.
- Dialogue and action get mixed together without collapsing into short choppy beats, and the rhythm varies line by line.
- Softened the omniscience rule. Characters can be all-knowing if the story actually supports it, which matters if you're playing gods or anything with a seer in it.
- 'Easy similes' is now 'unimportant similes', plus an explicit line against overwriting.
Prompt-level Tweaks
- Don’t Write for User: when a character directly addresses you and waits on a reaction, the scene ends right there. Should cut down on it answering for you.
- Formatting: added {{dialoguecolors}}. You need the Dialogue Colors extension for it, otherwise it just sits there doing nothing. I added it there because it's easier to track with the Macro option. Just remove it if you don't use it.
- Flexible length: now tells it to check the previous turns to work out how long the scene should be, instead of guessing fresh every time.
- Experimental Anti-Overthinking Prefill: added HTML guidelines to the checklist and made the permissive wording blunter, because some models were still hedging halfway through.
- Impersonation prompt: spells out that it's only your actions, thoughts and dialogue. It kept narrating for everyone else in the room.
Trackers and RPG bits (SillyTavern build only, the trimmed one doesn't ship these)
- Scene Tracker: the description between the tags can't be left empty any more. I kept getting hollow [SCENE] blocks with nothing in them.
- Status and Conditions Tracker: only truthful status effects, things that actually happened in the story. It was inventing conditions nobody had.
- Time Tracker: the hour has to be specific now, 3:47 PM rather than 'late afternoon', unless the setting genuinely makes the time unknowable.
- Persona-Based Stat Generator: utility skills are generated from your setting instead of the fixed Lockpicking/Analysis/Repair/Medicine/Survival list. That list was a leftover from when this thing was Fallout-flavoured, and I forgot to fix it because I'm a mouth open closing goldfish.
Settings and fixing typos
- Exported at 256k context, reasoning effort on high, media inlining off. These are just wherever my sliders happened to be, change them to suit your model and your patience.
- Typo fixes: 'charactets' in the Main Prompt, and Franz Kafka's name in the Bureaucratic Irony voice. That one had been wrong since I wrote it, sorry Franz.

15.0

Full Main Prompt rewrite... again, yeah. ~900 tokens or so without turning on anything else.
Grounded Prose Rules: wrote 'never' into it so GLM stops using them. Should work better now.
Fixes the bug where if you turn off something like NSFW mode, it still shows up cached in the main prompt (ST upstream bug), by using else statements.
Makes Optional User Instructions a macro as part of the Main Prompt instead when enabled.
Added the author names to the Narration Voices. May improve prose?
Renamed Prefills section to Prefills & Reasoning.
Added NPC naming rules as a separate toggle in Primary Toggles that also includes banned names (thanks to my friend Deigo for the list).
Added 'User Is Not A Character': new experiment; makes it so you are never a character in the RP and instead in the director's seat. Best paired with a persona named 'Narrator' or 'Director'.
Added 'Gossipy Voyeurism' Narration Voice more properly, which I previously didn't put in for whatever reason I forgot. Based on Bret Easton Ellis' American Psycho specifically.
Prompt-level Tweaks
- Formatting: removed the non-cache-friendly name randomiser macro. Now in a separate toggle in Primary Toggles section.
- Flexible length: ensures concise outputs. Turn this on to prevent overthinking in some LLMs while still keeping a reasonable length
- Gooner Mode: added that it is activated so that LLMs like GLM stop ignoring it by reasoning "well user didn't say it was activated lol".
- Reasoning Encouragement: moved to Prefills & Reasoning, added 'never draft' so it doesn't draft. Remove that part if you want it to draft.
- Experimental Anti-Overthinking Prefill: in my testing, makes Kimi K3 only think for about a minute. 300-1000 tokens of reasoning in my experience.
- Don’t Write for User: now uses 'never' wording so it stops trying to be funny and write for you.
Regex Fixes
- Makes it so regexes work globally no matter what.

14.0

Rewrote Main Prompt a bit again, 1200 tokens. Sorry chat. But the reason for this is that I folded in the Grounded Prose Rules a bit into the main prompt, theoretically you don't have to turn on Grounded Prose Rules ever. Maybe. Depends on your tolerance.
Grounded Prose Rules: expanded and reorganized into Core / Avoid / Use Instead; added many new banned clichés, romantic/possessive shorthand, audio formulas, environmental tropes, and explicit replacement guidance. This way it should make prose less dry while still hopefully banning things. Rule of Three will probably just become Rule of Four though, to be real with you.
Some toggles are now macros: Friction Mode, NSFW Mode, and Genre. This is so they are seamless to the system Main Prompt, makes it less confusing for the LLM. Also now you can see how much tokens they use on the main prompt in real time without doing mental arithmetic.
Prompt-level Tweaks
- Formatting: declared absolute override priority; added inner-thought rotation for multi-character scenes; exempted non-English honorifics from translation.
- Group nudge prompt: allows {{char}} to describe the actions, dialogue, thoughts, emotions, and perceptions of all characters present, including NPCs.
- Medium length: changed from 5–12 paragraphs to 8–13 paragraphs, excluding trackers and external HTML.
- Flexible length: now keeps outputs concise during regular character-driven scenes and slows down descriptively when the situation calls for it.
- Gooner Mode: added explicit priority statement and clarified it overrides the core directives' realistic world logic for the mode.
- Reasoning Encouragement: checklist now includes constraints, "never do X" wording, and director instructions.
- Experimental Anti-Overthinking Prefill: reasoning now uses a checklist of core directives and constraints.
- Don't Write for User: now excludes both <user> and the director. This will hopefully cause less impersonation. If it does it's probably the director framing, sowwy.
RPG ELEMENTS?!
- There are now three new prompts. They all go together and it's recommended to enable them all. Sorry for your wallet. I mean you don't need it anyway, this is just a gimmick, so don't blame me.
- Persona-Based User Stats: This will output first before anything else if enabled. It's based on the setting. At first I made it Fallout-esque, but I adjusted it to be more general.
- Level Up Companion: Levels up your level number, adds traits/perks based on how the story progresses and what you choose.
- CYOA with Skill Checks: Includes choices for you to pick with accommodating skills mentioned in the user stats. This prompt includes the dice roll element.

13.3

Rewrote the Main Prompt tighter, less contradictory, and more literary. It is now about 750-800 tokens depending on tokeniser.
Expanded Grounded Prose Rules to about 450-500 tokens and made it less dry when enabled.
Added Reasoning Encouragement for GLM doing fake thinking. It is not a custom CoT, just a small nudge for models that pretend they reasoned.
Made Don't Write for User and Write for User more concise.
Added Prose Voice options with Disabled as the default: Conspiratorial Absurdity, Bureaucratic Irony, Cosmic Playbook, Beige Undercurrents, Cruel Realism, Solemn Witness, Grand Satirical Stage, and Randomised.
Called it Prose Voice in-text so the model is less likely to start saying "the narrator" like it discovered a new toy.
Tried to make GPT 5.5 less wordy, or at least less like it ate Joss Whedon and Marvel movie scripts as its primary writing style source.
Strengthened tracker wording so enabled trackers should behave more consistently.

13.2

Main Prompt is now ~1200 tokens with a much stronger style, loosely influenced by Pseudonymous Bosch and Albert Camus. If the cheekier narration is not for you, remove the Narration Voice section in the Main Prompt.
Added a simplified ~600-token Main Prompt for smaller models and token-conscious setups.
Grounded Prose Rules is no longer required and is turned off by default. Turn it back on if familiar anti-slop issues start appearing.
Rewrote Friction Mode, Nightmare Difficulty Increase, NSFW Mode, and Gooner Mode to be more assertive and direct.
Adjusted Gooner Mode so GLM 5.1 is less likely to treat it as contradicting the main prompt.
Rewrote GPT Assistant Prefill to be less censored, though GPT 5.5 may still vary.
Main Prompt is now more neutral by default rather than always pushing a killer GM tone.
Added a Flexible Length Toggle, on by default. Short, Medium, and Long now use paragraph ranges instead of word counts.
Clarified tracker behavior so enabled trackers are considered consistently, with Pending Events limited to important events.

13.1

Made the Main Prompt more assertive about the actual style I want. It is now about ~900 tokens and has been working well in testing.
The stronger wording may increase negativity bias a bit, so reinforce nice characters through Optional User Instructions if needed. There is a filler instruction you can use there.
Added Length toggles: Short, Medium, and Long. Pick only one if you want a desired length, or choose none to keep length purely flexible.
The Main Prompt should mostly annihilate mouth open-close filler and further mitigate the "not a question" bullshit.
Added a pacing section to the Main Prompt to slow down GPT 5.5, though other models may remain proactive regardless.
Since I usually Write for User, I added and reworded anti-echoing instructions to mitigate character/user echo even further. It should be much sparser now, including on Claude and GLM 4.6/4.7 in my testing.
Rewrote and trimmed the Don't Write for User and Write for User toggles.
Added model-facing clarification for the Achievements Tracker and Scene Tracker toggles.
Switched Post-History Instructions back to Relative, System. It seems to adhere better when using None/Merge post-processing. Around ~400 tokens.
The style is stronger now, so read through the main prompt and edit it if you want a softer voice. My version leans blunt and sardonic on purpose.

13.0

Rewrote the entire Main Prompt. ~700-800 tokens
Trimmed Grounded Prose Rules slightly. ~400 tokens
Anti-Echoing is now part of Main Prompt, which should make models adhere to it better, including GLM 4.6/4.7, which I've tested. Sometimes Deepseek V4 Flash too.
Changed Grounded Prose Rules to injection for better adherence.
Instead of leaning too much into user being the director, it also equalises character adherence and autonomy, which causes better dialogue and characterisation in general.
GPT 5.5 actually writes like a person? Waow. Response length also isn't very high here. Make sure to set Post-Processing to None.
Added Experimental Anti-Overthinking Prefill for Kimi K2.6. Whether it works or not depends on factors like moon alignment, which gods are available this minute, etc.

12

Added Intimacy & Kink Randomiser Toggle.
Added "Direction Menu", which is situation-focused instead of user-focused choices to steer the plot.
Trimmed Grounded Prose Rules a bit.
Rewrote Write and Don't Write for User toggles to have stronger wording and anti-echoing.
Trimmed a lot of the rules of the toggles for the SillyTavern version, and gave them stronger wording.
Added two separate versions: the SillyTavern version which has all of the toggles, and a trimmed version which has only the Main, Primary Toggles, and the Prefills, for apps that ship these trackers and randomisers as agents.

11.5

Rewrote the entire Director Main Prompt to reduce redundancies and use words the LLM can use without bloating up tokens. ~800 tokens
Rewrote Grounded Prose Rules to be more consistent with Main Prompt. ~500 tokens
Rewrote both Write for User and Don't Write for User to enhance anti-echoing.
Removed 'Force Gemini to Think Less' now that it's irrelevant and replaced it with 'Friction Mode' - an extra toggle attempt to realistically combat positivity bias, especially in Opus. For some reason, Sonnet doesn't really have this problem.
Rewrote HTML Toggle to be more interactive.
Rewrote Difficulty Increase to be less bloated.
Rewrote NSFW Toggle with better wording.
Trimmed Parallel Off-Screen Tracker and World Detail Trackers and gave them stronger rules to follow.
Added "Pace" section in Scene Pressure Cocktail, added more random outcomes, and edited Combined Director's Cut in line with this.

11.0

Added two new trackers: Parallel Off-Screen Tracker, which shows what other characters are up to in ways that could converge with your plot without making your RP too deterministic; and World Detail Tracker, which includes small world-building details that could be used later in the RP for either conflict, resolution, or just enhancing the details of the world to make it feel more lived-in.
Added multiple Optional Randomiser Toggles:
- Grounded Complication → Moves the plot without it being too wild or making the plot too nasty.
- Chaos Mode → Moves the plot in bizarre, erratic ways without changing character behaviour arbitrarily.
- Genre Randomiser → Changes genres to inform what the scene will be about.
- Scene Driving Force → Picks a random type of main force that carries the scene forward (dialogue, action, reaction, tension, discovery, task flow)
- Scene Pressure Cocktail → Mixes several scene pressures together, combining tension, focus, consequence, and social weather into one blended scene texture.
- Combined Director's Cut → Combines all of the above in a way that still retains coherency.
- Dead Dove Escalation → Characters act on their worst impulses, but only when context supports it. Anti-softening. Try combining with Difficulty Increase toggle if Opus is still stubborn as hell about it.
As always, these are all optional. The RPs are still good without them, they just add texture. In my testing especially, Opus actually moves the plot well with these toggles.
Regexes are now prettier-looking.

10.7

In my attempts to fix Gemini 3.1 Pro Preview, I ended up overhauling the entire prompt. The good news - it makes every other model better, and it makes Gemini 3.1 Pro's dialogue sound less robotic. Bad news: The model itself is inherently nerfed. If I were Logan, I'm sure it would be the best thing ever by now, but I'm not. I'm just a guy.
Rewrote Grounded Prose to be more positive, so it should work on more stubborn models better now. You don't have to turn it off, but if it has problems with smaller models, you should.
My tears and soul are personally in this prompt.

10.6

Added more prose styling to make the environment livelier - it feels more like a scene you can imagine now rather than just pure functional dialogue and narration.
Moved main prompt up again. The GPT solution was simpler… use None post-processing.
Gemini dialogue writing should now follow your character card better rather than defaulting to its library of caricatures.

10.5

Added more director-esque framing into the main prompt since… you know, that's kind of the point.
Rewrote main prompt a bit to use more positive prompting, same with the writefor variables. Hence, now even Cydonia 24B won't write for you.
Restructured Main Prompt to be after Chat History. If it doesn't work as well for whatever model you're using, just move it back up to before Formatting.
GPT 5.4 should now adhere to main prompt directives better, so now it sounds less robotic and less like GPT trying to roleplay your characters like a bad actor on their first day of acting class.
Rewrote HTML Toggle to be more specific into what it can create.
Rewrote NSFW On to be what its purpose is - to cause mild NSFW framing
Rewrote Gooner Mode to be more hentai-ish.
Rewrote Difficulty Increase to be structured better, without contradicting itself.
Rewrote GPT Prefill to be stronger.

10.2

Finally added the README!
Removed the sentence macro, it was messing up some models.
Did a bit of rewording to make GLM 4.6 and 4.7 think if you don't have thinking turned off.
Rewrote Grounded Prose Rules a bit, it had some extra instructions that weren't really beneficial.
Placed the writefor variable back in Formatting, in case you want to turn off Post History Instructions.

10.1

Hotfix. A little bit of restructuring of prompts, so it's less of an eyesore.
Added an injection under Primary Toggles to 'Force Gemini to Think Less', for Gemini users.

10.0

Moved anti-slop into Post-History Instructions.
Better anti-echo and anti-repeating your words.
Gemini is now less of a horny freak and a jerk.
Better GPT 5.4 dialogue, shorter length.

9.5

Fixed Gemini's negativity bias by telling it to not rely on caricature.
Added something about secrets remaining hidden so characters put in more effort. Specifically this is a Claudism where a smart character keeps immediately guessing the ulterior motive within a message. Allows for more slow burn.
Rewrote default genre a bit.
Mitigated the "not a question" Claudism which pisses me off so much it's hitting me like a physical blow as I white-knuckle it.
Gemini is now not gonna default to horny mode unless you explicitly want it to be. So even if you have kinks or character states their penis size, Gemini is now not gonna try to have sex with you immediately if you don't want it to.
More naturalistic Gemini dialogue.
Rewrote anti-echo, anti-rhetorical echo. In my testing, even if you had it write for user for 50 messages, and you switch to don't write for user, it actually doesn't write for you. So that should hopefully mean it's effective.
Opus should still remain more neutral. Tested with some cards of sadistic people and cards that hate user and they don't soften immediately or try to do the typical Claude thing of "it's okay, we can therapy session immediately..."
Hopefully reduced caricature, exaggerated shorthand like "burnt water" or that stupid shit Claude likes to do.

9.0

Removed anti-slop injection in favour of tighter, positive prompting to improve style. So now it's transparent - the whole thing really is just 1100-1200 tokens, depending on which tokeniser is used.
Should reduce GPT 5.4 infinite yap works.
Some Claudisms can never be truly removed. This is the only thing constant in the world.
Removed redundancy and extra style paragraphs since they were pretty pointless.

8.5

Tightened style directives.
Removed XML Banned Patterns toggle (unnecessary).
Fixed regexes again.
Placed anti-purple prose as injection instead, alongside the banned patterns.
Reduced general redundancy.
Added a bit in both NSFW and Gooner Mode toggles.
More neutral Gemini 3.1 Pro Preview tone.

8.0

Fixed regexes.
Big style rewrite of the main prompt, so yes, unfortunately, it is now 900-1000 tokens. Sorry.
Added an XML version of Banned Patterns, which is now default, since most people who use this are probably using frontier models. If you are using models that prefer markdown instructions, you may disable the XML one and enable the markdown version.
GPT 5.4 fix, maybe. Better dialogue on that part, not sounding so stilted and "perfect". Use "GPT Assistant Prefill" for NSFW scenes. If you get refused, it's either my skill issue, yours, or OpenAI decided to be funny to taunt me specifically.
More decent prose on Deepseek V3.2, not having the "super short responses" problem, probably. GLM 5 also should have better prose and pacing adherence.
More neutral Opus - if a character really hates you, being nice to them most likely won't make them like you immediately. Again, this depends on how you write the card.
To clarify, Difficulty Increase is supposed to make the difficulty ridiculously difficult. It is not an anti-positivity bias. Turn it on if you want silly-fficulty.
Theoretically, should still be customisable and easy to put your extra stuff into. Make sure to read the main prompt to see which parts you'd like to modify/remove for your own preference. The writing style indicated here is one I vastly prefer on my own, maybe you don't like it, maybe you do.

7.5

Since the anti-slop is getting bigger, I significantly trimmed the main prompt.
If you aren't using a SOTA/recent model, the anti-slop is probably unnecessary, so you can turn it off.
Makes sure GLM doesn't always end on a question.
Just uh, better writing in general in my opinion.

7.0

Total rewrite of main prompt and anti-slop
Significantly less Claudeisms
Less purple prose, less unhinged melodrama (looking at you, Claude)
More emotional Gemini 3/3.1, but without being overwrought
Encourages models to be more neutral

6.0

Trimmed main prompt to save tokens (again)
Rewrote Difficulty Increase prompt so you can theoretically turn it on forever, but still have the characters get hard-earned character development and softening
Fixed all regexes
Ensured Opus 4.6 doesn't yap forever

5.7

Rewrote the whole main prompt
Responses now respect actually ending on a hook when writing for user
Added Time Tracker, Character Status Tracker, Secrets Tracker, and fixed Reputation Tracker and improved Relationship Tracker
Clarified Difficulty Increase to Nightmare Difficulty Increase

Using Pura 15.1 on my Kaveh & Alhaitham card with the same greeting scene. I used twelve different models. I do this whenever I change something big in the main prompt. Reading the diff back doesn't tell me how the writing actually comes out.

These aren't benchmarks, I just wanted to show some transcripts on how I usually roleplay and what the prose looks like. It's unedited, I kept the stupid mouth open closing too. Grounded Prose Rules prompt is off for this entire pass.

Most of them run Don’t Write for User. The last one is Write for User on purpose. I'm usually lazy so I actually co-write most of the time, but hey, you can see I actually test my preset a lot!

In my opinion, prose matters a lot to make narration not exhausting to read. The 'pass' stuff here is just for LOGIC errors. I like Muse Spark's prose and dialogue a lot, for example, and I find I get Opus fatigue pretty fast.

Opus 5 Pass

Makes the whole thing fun to read and still doesn't answer for me. Everyone sounds like a different person. It does like inventing small world details.

Read the transcript →

Kimi K3 Pass

Definitely my absolute favourite, it picks things up from earlier to use against people later. Also, great prose and voice differentiation.

Read the transcript →

Fable 5 Pass

Best one for agency. Hands me the decision and the NPCs still push back. Heat and time of day change how people act. Still pretty dry in my opinion though.

Read the transcript →

DeepSeek V4 Flash Pass with caveats

A budget option that follows the prompt so-so depending on its mood.

Read the transcript →

GPT-5.6 Sol · Medium Pass

Sol always wants something to happen. 15.1 doesn't stop that, it just aims it somewhere sensible.

Read the transcript →

Opus 4.6 Pass

Good at who knows what, and at people acting against how they'd describe themselves. Explains the point a bit too soon.

Read the transcript →

Muse Spark 1.2 Pass

Big improvement for a small model. In my older tests Kaveh would randomly start nyaing for some reason, and that's gone now. It's still very sure about things it guessed.

Read the transcript →

GLM 5.2 Pass with caveats

Behaviour improved, the tics didn't. Still the most stubborn model I own for saying my own words back at me.

Read the transcript →

Gemma 4 31B Pass with caveats

Unfortunately it explains everyone to death, but it's still a very solid model with voice differentiation and doesn't really have basic logical errors up the ass.

Read the transcript →

Gemini 3.6 Flash Pass with caveats

Great dialogue, then it picks me up and moves me through the market without asking. The one clear Don’t Write for User failure here.

Read the transcript →

Mistral Medium 3.1 Mixed

Much better than the last time I tested it, and it still forgets boundaries it set two paragraphs ago.

Read the transcript →

Kimi K3 Pass Write for User

How I usually roleplay. I have a persona that's about 600 tokens, yes.

Read the transcript →

Overview Features Prompts Agents Included Tools Screenshots

Neconyan

v1.0.0 · The Cutest LLM RP Frontend™

My own SillyTavern fork, with its own interface, defaults and direction. Character chats, writing tools and model settings all live in a cat-themed workspace that works on a phone as well as it does on a desktop.

I run it on a small server and use it mostly from my phone, so phones aren't a squashed desktop here. Anything you can reach on a desktop you can reach on a phone in the same number of taps, and there's no simple/advanced switch hiding half the settings from you.

It keeps SillyTavern's file formats for characters, chats, lorebooks, personas and presets, so bringing your stuff over is a starting point rather than a rebuild. It isn't chasing upstream parity though - features, layouts and behaviour can differ, and extensions that depend on SillyTavern's page layout may need changes.

Four modes: Roleplay, Conversation, Meower and Story Mode.
Pura's Director Preset 16.0 and The Ethereality Express 1.1 come bundled as the default presets. The trackers and randomisers ship as agents instead of preset toggles.
Replies keep generating on the server if your phone kills the tab, and land in the chat when you come back.
Mewmory keeps track of what happened in a roleplay, and you can look through it, correct it or undo it when it gets something wrong.
There's an Android app now. It runs the whole thing on the phone itself, so no Termux and no separate server - you just bring your own model connection. It needs Android 11 or newer and Android System WebView 124 or newer.
Default address is http://127.0.0.1:4433/. Bun or Node.js, with launchers for Windows, macOS, Linux and Termux.

Prompts

Pura's Director Preset / Neconyan

This is the plain preset Neconyan ships with - no trackers, those are agents in Neconyan. Pick one to read or copy; the text comes straight from the current preset. Placeholders and macros are unchanged, so some prompts still need a compatible roleplay app to work.

Original prompt (with app macros)

What's Different

The stuff I actually wanted out of a frontend

Workspace

A sidebar that stays put on desktop and slides in on a phone, with search built in. Your characters, recent chats and tools are one tap away instead of buried in nested drawers.

Four Modes

Roleplay is the classic chat. Conversation turns a character into a messenger-style DM with a Pals rail. Meower is a social timeline, and Story Mode is for long-form writing.

Phones Come First

Labelled tab rows instead of dropdowns, 44px touch targets, and a sign-in that can remember your device for 30 days. If your phone's browser kills the tab mid-reply, the server finishes the reply anyway and it's waiting in the chat when you come back.

Mewmory

Long-term story memory with its own model roles, source-backed NPC references and interview histories. It can run on a cheaper model than the one writing your replies.

Assistants and a Tour

Miso, Taro and Nori are the three built-in assistants, each with male, female and neutral variants, and there's a short First paws tour for your first launch. You can hide it and bring it back from the Home layout.

Calico Everything

Calico themes, an animated cat, local fonts and a pile of backgrounds. Every bundled theme is checked for readable contrast, and animations respect reduced motion.

Your Data Stays Yours

Same file formats as SillyTavern for characters, chats, lorebooks, personas and presets, and no telemetry. Back up your old install and use a separate data folder when you move over.

Agent Stack

Small models doing the boring work so the good one can write

In-Chat Agents

An agent is a stored prompt with its own conditions, stage, connection profile and output handling. Toggle them, group them, import and export them. Each one can run on a different model than your main chat, which is the part I use most - a cheap model doing tracker work while the expensive one writes prose.

The Pipeline

Agents can run before the reply to add rules or trackers, alongside it for commentary, or after it for a second pass over the finished text. The after-reply pass is where proofreading and prose polishing happen.

Companion Agents

Companions make their own model calls and store the results instead of touching your reply - that's how the tracker panels get filled in. They can run all at once or one at a time. Parallel if you're impatient like me, sequential if your provider starts complaining about rate limits.

Built From the Presets

Every bundled tracker agent uses the same prompt and regex as its toggle in Pura's Director Preset 16.0 (SillyTavern) or The Ethereality Express 1.1, so the cards look the same either way. That's why the Neconyan preset can stay lean.

Proofreader by Default

My Proofreader agent comes installed, switched off until you want it. It runs after the reply and rewrites the finished text as a second pass, so the main model doesn't have to get everything right in one go.

What Ships Right Now

Every Pura tracker and randomiser, the Ethereality Express trackers and Doors & Dice menus, formatting helpers, content toggles, preset groups, and companions like Expressions and Plot Compass.

Included Tools

Things that come in the box

Agents

Helpers that run before or after a reply. Trackers, randomisers and the Proofreader live here.

Mewmory

Long-term memory with its own model roles.

Chat Archive

Keeps old chats tidy and findable.

LoreStitch

Lorebook tools.

BotSearcher

Search supported character-card sites and import cards straight into Neconyan.

Quick Image Gen

Generate images from inside a chat.

Guided Generations

Steer the next reply with a short instruction.

Input History

Bring back what you typed earlier.

Prose Polisher

A second pass over the finished reply.

UI Preview

Real screenshots from 1.0.0, on a desktop and on a phone

Desktop

Home

Roleplay

Conversation

Agents

Phone

Home

Roleplay

Conversation

Overview Options Prompts Trackers Sample Older Versions

The Ethereality Express

Latest Version: 1.1

The magical realism sibling of Pura's Director Preset. The impossible is mundane, nobody explains it, strangeness follows emotional logic, and the prose stays flat and direct. The total opposite of the grounded realism in the director preset. Main prompt is roughly 1300 tokens.

This is specifically made for a world that literally lives - so expect strange things to happen, such as objects talking to you, being followed by fruits, and so on. The style is written to be as strong as possible to push the model into hard magical realism. Turn on Anchor mode if you'd like a break from all the weird stuff. I tried to make this preset as easy to edit as possible; hopefully you feel that you can edit it as you like.

New in 1.1: Pressure Modules (Friction and Nightmare Difficulty) that stack with any Veil Mode, six new Chance Events, two Scene Dice, two new narration voices, and a Reasoning Encouragement checklist. The Ledger is called Owed to the Strange now, the Department of Small Miracles issues proper paper certificates, and The Committee keeps its roster out of sight.

By default, these are on: Nightmare Difficulty, Unscheduled Phenomena, The Story Remembers, Drift, Omen, The Clock Is Lying, The Census, Doors and a randomised voice. Everything else in Preset Options is off until you switch it on.

Veil Modes set how much strangeness a scene is allowed. Toggle at most one; with none on, the main prompt's own default level applies.

More notes

Trackers are all optional. Each one makes the model print a small labelled block of text, which a regex then renders as an animated card. Toggle whatever you like, but running all fifteen at once means fifteen sets of formatting rules competing for the model's attention, and output quality drops well before that. Poor model…

While I tried to make it compatible with NSFW and intimate scenes, your mileage may vary, as always. It's possible it won't mesh with the average person's idea of intimacy; after all it's supposed to be combined with magical realism.

In my experience it seems to ensure the prose is less sloppy, however it is more metaphorical. If you find descriptive prose to be automatically purple, this isn't for you.

Tested on Kimi K3, GLM 5.2 and 5.3, Deepseek V4 Pro and Flash, Gemini 3.7 Flash, Gemma 4 31B, Gemma 4 26B A4B, Gemma 4 12B (ran locally), Opus 4.6 and 5, and Fable 5. Still need to do further testing for other usual ones, but so far so good.

Inspirations: Authors in Narration Voices and Disco Elysium.

Prompts

The Ethereality Express

Original prompt (with app macros)

Preset Options

What each toggle actually does

One file for SillyTavern. 1.1 ships ready to run: the Liminal Main Prompt, Genre, Formatting, Ethereal Prose Rules, Reasoning Encouragement, Write for User and Flexible length are on, plus Nightmare Difficulty, Unscheduled Phenomena and The Story Remembers, a randomised voice, Doors, and four trackers - Drift, Omen, The Clock Is Lying and The Census. Everything else is off until you switch it on. Grab the regex set from the Preset tab too, or the trackers arrive as raw brackets.

Toggle whatever you like, but running all fifteen trackers at once means fifteen sets of formatting rules competing for the model's attention, and output quality drops well before that.

Core Prompting

Liminal Main Prompt

The whole preset, around 1300 tokens. You're the overseer, impossible things are ordinary, and characters treat the strange the way you'd treat a leaky tap.

Ethereal Prose Rules

The post-history anti-slop layer, which bans the word 'ethereal' in a preset called The Ethereality Express. Leave it off on smaller models.

Genre

Two sentences telling the model what magical realism actually is, mostly so it stops staging strangeness as a big reveal.

Formatting

Present tense, third person, dialogue in quotes, no asterisk narration. Delete the {{dialoguecolors}} line at the bottom if you don't use the Dialogue Colours extension.

Length

Short is 3-5 paragraphs, Medium is 8-13, Flexible reads the scene. Pick one; I run Flexible.

Write for User / User Controls User

Pick one. Write for User is on by default. User Controls User replaces 1.0's Don't Write for User: your actions, lines and thoughts stay yours, and the reply stops at the first decision that needs you.

Optional User Instructions

An empty slot for whatever you want to add on top. There's a little easter egg line in there, so clear it out first.

NSFW

The same flat, physical register as the rest of the preset, with the strange turning up in it and leaving something behind. If that isn't your idea of intimacy, leave it off.

Reasoning Encouragement

New in 1.1. A silent checklist before each reply: your instructions and canon first, then the Veil budget, then pressure, dice and trackers. It also sorts out what happens when two Chance Events want the same reply.

Veil Modes (one at most)

Veil Thin

Two or three strange things a scene: rooms rearranging by mood, doors leading somewhere adjacent. The loudest of the five.

Anchor

One small strange detail per scene at most, with the rest spent on chores, money and people being irritated. I use it to cool down after something big.

Elegy

Strangeness through memory and absence, like a coat that keeps its owner's posture. Grief stays practical.

Feral Wonder

Folk-tale rules, so the impossible bargains, keeps score and sends a bill. My favourite of the five, honestly.

Farce

The ridiculous played completely straight, all committees and paperwork for miracles. The stranger the event, the drier the sentence.

Pressure Modules (stack with any Veil Mode)

Friction Mode

New in 1.1. People get harder to convince. They guard their time, pride and version of events, every refusal has a reason, and every concession comes with a changed condition.

Nightmare Difficulty

New in 1.1, and on by default. Eldritch horror that runs on a consistent local rule: you get a warning, a pattern and something you can do before it escalates, and failure leaves a lasting cost.

Narration Voice

Generational Murmur

Márquez. Long, warm sentences that treat decades and dinner with the same patience.

Quiet Static

Murakami. Cool and plain and faintly detached, anchored in cooking and laundry, with the strange arriving sideways.

Lucid Geometry

Calvino. Playful precision, rooms and relationships handled like elegant machines with rules.

Velvet Teeth

Carter. Rich, sensuous and a little dangerous, where desire and hazard share the same detail.

Documentary Dread

New in 1.1. Measured, exact sentences and a bit of archaic vocabulary, with dread built out of evidence and scale.

Humane Absurdity

New in 1.1. Quick, dry and kind. The jokes go after institutions, greed and foolish systems, never at grief.

Randomised

Rolls one of the six each reply. Chaotic, and I enjoy it more than I expected to.

Chance Events (choose up to two)

Unscheduled Phenomena

A silent d20 every reply decides whether reality amends itself partway through the scene, and a natural 20 brings a second amendment later in the same reply. The threshold sits in the prompt as a plain 14, so raise it for a calmer world.

Ordinary Trouble

New in 1.1. A d6 where 5-6 brings one completely mundane complication - something breaks, time runs short, the door is locked. It doesn't spend the Veil's weirdness budget.

The Story Remembers

New in 1.1, and on by default. On 16-20 of a d20, one loose end you already left lying around comes back: a promise, a rumour, an unpaid hour, an unfinished job.

Small Mercy

New in 1.1. On a natural 20 the world helps once, small and oddly specific, and it really is free. You still get to accept, refuse or question it.

Collection Day

New in 1.1. On 17-20, one debt or promise on record comes due on its original terms, with a warning and at least one way to pay, argue, stall or run.

Wrong Address

New in 1.1. On 19-20 something meant for somebody else turns up - a parcel, an invoice, a curse, a minor miracle - and what you do with it becomes canon.

Borrowed Luck

New in 1.1. A natural 1 or 20 moves luck between pockets. Your disaster is somebody else's windfall, preferably someone you don't like, and the other way round.

Scene Dice (choose up to two)

Scene Engine Die

New in 1.1. A d8 picks what carries the reply: dialogue, action, aftermath, a practical problem, social pressure, discovery, tension, or a shared task.

Social Weather Die

New in 1.1. A d10 gives the NPCs a mood for the scene - everyone's tired, someone wants out of the conversation, someone is in a better mood than the scene deserves.

Trackers

Drift

A needle measuring how far the scene has moved from ordinary, MUNDANE up through to SUBLIME. It only files when the level genuinely changes.

Motif

Counts recurring symbols and what they've picked up: objects, animals, phrases, weather. It logs continuity rather than meaning, or the model starts writing an essay about itself.

Omen

Signs the world plants about what's coming, DISTANT through to ARRIVED or PASSED OVER. Readings are allowed to be wrong.

Thin Places

Locations and hours worn through at the elbow: the pantry, the sixth stair, 3:03 AM. Once a thin place has rules, those rules stick.

Absurd Trackers

Entanglement

Relationships as physical objects, like a red thread at the wrist or a debt of three teaspoons of sleep. One form per pair.

The Clock Is Lying

Official time against actual time, both correct at once. It says 3:47 PM, the truth is 'two griefs past noon'.

Improbable Effects

Inventory, except the possessions have moods and demands. Anything borrowed from the strange accrues interest.

Afflictions & Blessings

Impossible conditions people catch, inherit or earn, like casting two shadows, one of them punctual. Cures always cost something specific.

Owed to the Strange

Debts owed to and by the strange: three true compliments, one afternoon of childhood, the good scissors. It always collects, always slightly askew of the terms. Called The Ledger in 1.0.

What the Town Knows

Rumours kept like livestock, hatching and feeding and mutating. They drift towards the satisfying version rather than the accurate one.

Department of Small Miracles

An unseen bureaucracy issues deadpan certificates for notable impossibilities, form numbers and office hours included. One per scene at most. 1.1 prints them on aged paper with a wax seal and a stamp.

Meanwhile, Impossibly

Three bullets of off-screen strangeness - an absent character, the town, and something owed or growing - tagged with how close it's drifting towards you. 1.1 flips a coin each reply and only prints when there are three facts already established to copy, so it records the world rather than inventing it.

The Thought Cabinet

Straight out of Disco Elysium. Ideas move into a head as tenants, then either settle in permanently or get evicted and leave a dent behind.

The Committee

Also Disco Elysium. Four to six named inner faculties interrupt mid-prose with a pass or a fail, and a failure stays confidently wrong until events correct it. Mostly fluff; it isn't tied to any skills or numbers.

The Becoming

The story stitches people an epithet out of whatever they keep doing. The Apology Landlord. The Tuesday Widow.

The Census

Registry-style profile sheets for new named characters, sorted into RESIDENT, LODGER and PASSERBY. Nobody is ever seen collecting it.

Doors & Dice

All off by default. Two regexes in the set strip old Doors and Winds menus out of what gets sent to the model, so they don't pile up in your context; the copies already in your chat log stay visible where they are.

Doors

A choice menu of 3 to 7 options at the end of every response, some treating the impossible as an errand and some entirely mundane. One is always slightly ill-advised.

Doors with Fate Checks

The same menu, except every option carries a competency and target number from your Almanac and is resolved by one hidden d100 whose arithmetic never reaches the page. Pick this or plain Doors, never both.

The Four Winds

Four plot pivots at the end of every reply, two grounded and two where the impossible arrives at the worst moment. Sits fine alongside Doors.

Almanac Generator

Rolls you a stat sheet from your persona and history: Constitution, competencies, Traits, Knacks and Burdens. It prints once at the start when there's no Almanac yet, rerolls if you ask, and works from the existing sheet after that. Still best switched off once it has printed.

The Turning

Level-ups, kept rare. A season only turns after something that mattered, so ordinary wins accumulate in silence and get nothing.

Unscheduled Phenomena

Improbable Effects

Afflictions & Blessings

Owed to the Strange

What the Town Knows

Department of Small Miracles

Meanwhile, Impossibly

Doors

The Almanac

One run of 1.0 on my Nagito Komaeda card. The defaults were left alone; Drift, Omen, The Clock Is Lying, The Committee, The Becoming and The Census, with Unscheduled Phenomena switched on as the only addition and no Veil Mode picked.

The trackers below are rendered by the regex set exactly the way they land in the chat, so this is also the easiest way to see what the cards actually look like before you install anything.

GLM 5.3 Nagito Komaeda

Room 13 and the Warm Keycard. A pre-dawn beach, a keycard that stays warm to the touch, and a hotel corridor that agrees to exist. Drift, Omen, Hour, Voice and Unscheduled Phenomena all fire across the three turns.

Read the transcript →

Previous Preset Versions

Older releases

The Ethereality Express 0.5