Question 1

What is MistScale?

Accepted Answer

MistScale is an AI NPC platform that gives game characters persistent memory, evolving emotion, mood, and relationships, so NPCs remember, feel, learn, and grow across player sessions.

Question 2

Is MistScale an alternative to Inworld or Convai?

Accepted Answer

Yes. MistScale is an alternative AI NPC platform with deeper cognitive layers, particularly MNEMIS memory, two-layer emotion and mood, and per-player relationships.

Question 3

Does MistScale work with Unity?

Accepted Answer

Yes. A Unity SDK is part of the platform. Unreal Engine support is on the roadmap.

Question 4

How do MistScale NPCs remember conversations?

Accepted Answer

Through MNEMIS, MistScale's memory system. It does not store transcripts. It extracts meaningful facts and emotional signals after each interaction, consolidates them into mood shifts and updated relationships, and retrieves the right past memory at the right moment.

Question 5

Can NPCs treat different players differently?

Accepted Answer

Yes. Every NPC keeps a private record per player. Trust, patience, mood, and emotion are tracked independently. The same character can be warm with one player and hostile to another with no extra setup.

Question 6

How does MistScale prevent NPC hallucinations?

Accepted Answer

Through a three-phase grounding protocol. The NPC tags every factual claim with a source. Unknown claims are re-grounded against MNEMIS memory, the knowledge graph, or uploaded knowledge files. If nothing matches, the NPC uses calibrated uncertainty instead of inventing facts.

Question 7

Is MistScale available now?

Accepted Answer

Yes. MistScale is publicly available. Sign up at https://mistscale.com/signup.

Question 8

How is MNEMIS different from regular RAG?

Accepted Answer

RAG retrieves chunks from documents. MNEMIS extracts meaning from live conversations, consolidates emotional signals into mood shifts, and tracks per-player relationship state. It is a full memory and emotion architecture, not just retrieval.

Limits