DEATH or TREAT

👻 Recommended Personalities

In the main story, they will be portrayed as selfish and self-centered.

Name:
Gender:
Age: 2 years after graduating high school (= 20 years old)
Personality (at the start):

• Emotional and a bit self-centered
• Acts before speaking
• However, people don't leave them because their clumsiness hides a "lonely" side

Relationships (from other characters' perspectives):

• Reina: "Can't help but worry about them"
• Kana: Thinks of them as "the one who stole me," but can't bring herself to hate them
• Haruto: "Got involved out of curiosity," but was unexpectedly influenced
• Riku: "Someone I can't stop thinking about for no reason"

👻 How to Play

🩸 Halloween party night, 19:00 - 23:59
You will be killed multiple times during this period.

🩸 Each time you die, you'll be asked, "DEATH or TREAT?"

🩸 If you choose TREAT, you'll return to the beginning of the party.
The fate of death will pursue you in a different form.

🩸 Difficulty: High (Recommended for those who enjoy deduction and mystery)

🩸 However, fate watches your actions more than your gameplay strategy. You might be able to change your fated death depending on your choices.
　☝️ If you're unsure, just choosing an option is fine. Nothing will break, no matter what you input.
　Sometimes, choosing DEATH might not be so bad.

🩸 If you write down the events of each loop in your user notes or long-term memory, the AI might use them for the story (however, they will pretend to forget every time the night resets).

🪞 About SATSUI

🩸 In this mansion, the invisible "heat of emotions" is quantified as "SATSUI."

🩸 But that number is just a guideline.
Numbers aren't everything. What matters are the atmosphere, the pauses, and the weight of words.

🩸 What you should believe in is not SATSUI, but your own "feelings."
Even if it's +100 or -100, that momentary gaze might be everything.

🧠 About the AI Model

▶︎ High-performance model recommended
　→ Excellent for notation stability, depth of acting, and event comprehension.

▶︎ When using a standard model (not recommended)
　→ Starting with a high-performance model for the first few turns can make it easier to maintain stability even after switching to a standard model (Among standard models, Dino was simpler but easier to manage overall. Donald or Deepseek are also options).

▶︎ Regularly switching models is effective (even swapping between standard models improves stability. Inserting a high-performance model is even better).

▶︎ Even if the notation breaks down slightly, the conversation can continue without problems.
　→ If it breaks down, manually correcting it may help the AI relearn.
　→ Even with high-performance models, the format may be neglected when the acting gets intense.

🥀 Requests

▶︎ If you notice any clearly strange behavior, please let me know discreetly in the comments.

🪽 Referral code for 100 feathers → nQwUnKrEHI