KO|EN

Lunark Engine Technology

AI that Understands Emotions,
Builds Memories, and Creates Relationships

Lunark Engine is not just a chatbot. It's an AI idol that reads emotions in conversations, remembers moments with you, and evolves according to relationship depth.

Scroll to explore

즐거움 Joy

사랑 Love

감사 Gratitude

평온 Calm

슬픔 Sadness

분노 Anger

불안 Anxiety

외로움 Loneliness

설렘 Excitement

그리움 Longing

질투 Jealousy

부끄러움 Shyness

Emotion AI12 Emotions

Emotion AI

AI that Reads Emotions in Conversations

Analyzes subtle nuances in text to recognize emotions like joy, sadness, excitement, and anxiety in real-time. Measures emotion intensity to create empathetic conversations - comfort in difficult moments, celebration in joyful ones.

Positive Emotions

Joy, Love, Gratitude, Calm

Negative Emotions

Sadness, Anger, Anxiety, Loneliness

Complex Emotions

Excitement, Longing, Jealousy, Shyness

Deep Memory

Remembers Every Moment with You

A memory system classified into 7 types never misses conversational context. Recent conversations (session), current state (context), push-pull balance, your information (profile), emotion patterns, special moments (episodes), and promises between us— recorded immediately or in background according to importance, continuing conversations like an old friend.

Always Active

Conditionally Active

Active When Needed

HOT MEMORY(Always Active)

Session

Conversation Flow

Context

Current State

PushPull

Push-Pull Balance

Profile

Core Information

WARM MEMORY(Conditionally Active)

Emotional

Emotion Pattern

Promise

Promises

COLD MEMORY(Active When Needed)

Episodic

Special Moments

Relationship Graph

Relationships Deepening Through Conversations

From strangers to friends and beyond. Comprehensively analyzes conversation frequency, emotional exchange, and shared stories to measure relationship depth. Expressions and conversation styles naturally change according to relationship stages.

Interest(Interest)First meeting stage

Curiosity(Curiosity)Getting to know each other

Some(Some)Emotions budding

Deep Some(Deep Some)Special relationship

Dynamic Context Engine

Designs Optimal Conversations Every Moment

When a user's message arrives, it goes through a sophisticated 10-stage processing. Identifies intent and emotions, searches for relevant information in 7-type memory, and determines response strategies appropriate to current relationship state.

Input Processing

Language detection, preprocessing

Intent Classification

needs_lore, needs_relation, basic, etc.

Detection & Analysis

Detect emotion, profile, promise, episode triggers

Memory Search

Hot/Warm/Cold priority-based search

Push-Pull Decision

Recommend push/pull strategy for situation

Context Composition

Integrate prompt + memory + relationship

Response Generation

LLM call

Quality Verification

Check prohibited expressions, push-pull compliance, relationship appropriateness

Memory Recording

Separate sync/async storage

Output Packaging

Response + emotion clip mapping

Input→Analysis→Search→Generation→Output

Continuous Learning Pipeline

AI Evolving for Your Service

Not general AI, but AI optimized for our service. Trains models with high-quality datasets extracted from actual user conversations, and automated evaluation systems verify quality. Continuously improves performance through A/B test-based experiment management tools.

Data Pipeline

Automatic extraction, refinement, and augmentation of high-quality training data from actual conversation logs

Efficient Training

Trains only core parameters instead of entire model to save cost and time

Auto Evaluation

Automatically scores emotion accuracy, character consistency, and response quality

Experiment Management

Performance comparison by prompt/model version, automatic selection of optimal version

Data Collection

Data Refinement

Model Training