Presets
Chrysanthemum Preset
SillyTavern Preset

Chrysanthemum 2.0.0

A simpler sibling to White Lotus, built for one-on-one RP and characters with agendas of their own.
chrysanthemum-latest.json
v2.0.0  ·  Latest  ·  SillyTavern Preset  ·  JSON
Download
Changelog  |  Previous: v1.0.0

Chrysanthemum is an alternative to my White Lotus preset. I liked how White Lotus was structured, but I wanted something simpler — so this is that. Same general goal, less machinery.

2.0 switches from markdown to XML and puts most of the character instructions in a Contract. The focus is on characters behaving like people: having their own priorities, getting things wrong, keeping things to themselves, and doing something besides waiting for {{user}} to tell them what happens next.

It doesn't prescribe a particular author's style. By default, narration is third person, past tense, limited to what {{user}} can perceive, with a tone suited to the POV character. It's still built for one-on-one roleplay, and it doesn't require the Experimental Macro Engine.

Characters With Their Own Priorities

The Contract covers character autonomy, relationships, knowledge boundaries, and keeping the world physically consistent. Characters can be selfish, mistaken, or unhelpful. Being close doesn't mean they suddenly explain all their feelings, and {{user}} isn't automatically the center of everyone's world.

Dialogue takes its vocabulary from the setting and time period, and characters can speak differently depending on who they're talking to. The goal is to move the scene through their own choices, without every turn becoming polished banter or a recap of what you just wrote.

The Toggles

NSFW On adds the explicit-content instructions. Write for User lets the model write {{user}}'s actions, thoughts, and dialogue too; leave it off if you're playing your own persona. This replaces the old User Autonomy switch, so the direction is reversed: turn this one on to hand over control.

Length has Short (at most one paragraph), Average (2–4 paragraphs), Long (7–10), and Flexible (whatever the scene needs). Average is selected in the download. Enable only ONE length option at a time — turn off the current one before choosing another.

Guidance contains prose examples to steer the model away from common annoyances. Turn it off if you'd rather save those tokens. There's also an optional Jailbreak, originally made for MiMo 2.5; I can't promise the same results on other models. YMMV.

Tweaking It

The Contract holds the characterization rules, Speaking Style handles dialogue, and Turn Start handles tense, formatting, and how the next reply continues the scene. The built-in Main Prompt slot is empty and off by default. Leave Reset Variables enabled so the options reset correctly when you change them.

I primarily use Kimi, GLM, and MiMo, so those are the model families this was tested on. These are the settings saved in the download; adjust them for your model and context needs.

Included Settings
Temperature 1  ·  Top P 0.95  ·  Top K 40
Context 50,000  ·  Max response 5,000 tokens
  1. 01Download the JSON and import it as a Chat Completion preset in SillyTavern.
  2. 02Check your model and context settings, then choose a length and the options you want.
  3. 03Leave Write for User off to keep control of your persona. Load your character and go.