The Magic of Fear.

When we talk about AI in games, we don’t really mean artificial intelligence.
What we mean is behavior — carefully authored illusions designed to feel intelligent, unpredictable, and scary.

For this horror project, the goal isn’t to build a perfect thinking machine.
It’s to control tension, pacing, and player psychology while making the enemy feel aware and alive.

So instead of going full GenAI or overengineering with systems that aren’t needed yet, I’m building a modular, director-driven enemy behavior system that fakes intelligence convincingly.

Below is a breakdown of how the current system works.

1. The Director – Controlling Fear, Not Just Enemies

At the top of the system is the Director.

The Director doesn’t move the enemy or trigger scares directly.
Instead, it controls pacing and tension:

• Prevents constant jumpscares

• Manages buildup → peak → release cycles

• Decides when the enemy is allowed to act

• Ensures the player gets breathing room after intense moments

Every action the enemy wants to perform goes through the Director first.
If the tension budget allows it, the action happens.
If not, it’s denied or delayed.

This keeps the experience scary without becoming exhausting.

2. The Agent – Actions, Illusions, and Feints

The Agent is the enemy itself.

Rather than hardcoded logic, the enemy is built from Action ScriptableObjects:

• Move

• Teleport

• Appear

• Chase

• Vanish

• Fake-outs (feints)

Each action has:

• An intensity curve that defines when it’s allowed (based on tension)

• Parameter curves (speed, distance, duration)

• Authorable values for fine tuning

This makes behavior feel dynamic without randomness chaos.

Feints & Deception

The agent can intentionally lie to the player:

• Start a chase, then vanish

• Appear from one direction, reappear from another

• Trigger movement sounds without visual confirmation

The goal is to make players question what they saw, not just react to it.

3. Emitters – The Player Is Always Talking

The player constantly emits information.

These are handled through Emitters, such as:

• Idle Emitters – standing still for too long

• Area Emitters – entering or staying in risky zones

• Trigger Emitters – interacting with objects

• Mic Emitters – making noise

Emitters don’t force actions.
They simply inform the system that the player did something.

The agent asks the Director:

“Can I respond to this?”

If the budget allows, the enemy reacts.
If not, the system waits — creating delayed fear and unpredictability.

4. Scare System – Controlled Chaos

Scares are handled separately from core behavior.

The Scare System:

• Randomly selects from authored scare events

• Triggers visual, audio, or movement-based scares

• Avoids repeating patterns

• Is gated by the Director to prevent spam

Scares can include:

• Sudden appearances

• Lunges toward the player

• Disappearing when looked at

• Environmental sounds and movement

The key idea:
Scares should feel earned, not constant.

Design Philosophy

This system is:

• Modular

• ScriptableObject-driven

• State-machine based

• Easy to tune via playtesting

Behavior Trees or GOAP may come later, but right now the rule is:

If it’s not needed, don’t do it.

The illusion matters more than the complexity.

This is still early, and a lot will change as playtesting continues — but the foundation is already flexible enough to grow without breaking.

Upcoming Features

Custom Mode

A fully customizable mode where players can tweak:

• Time of day

• Number of notes to collect

• Enemy aggression and behavior tuning

Designed for both accessibility and challenge runs.

VR Mode

A dedicated VR mode is planned:

• Reworked pacing for VR comfort

• Adjusted scare timings

• Presence-first enemy behavior

Fear hits differently in VR — and the system is being built with that in mind.

More dev logs soon as the systems evolve.
Thanks for reading, testing, and supporting a solo dev trying to scare you responsibly.