Product Highlights
Emotion Recognition
The emotion recognition model can accurately identify the emotional states of both the AI and the conversational partner. By referencing character settings and desires, it generates an emotion vector and an appropriate emotional response.
Desire System
Based on the AI character's goals and current situation, desires are generated. Using the Monte Carlo Tree Search's inference structure, it searches for the path to realization.
Memory Scheduling
Self-awareness, rational memory, emotional memory, knowledge graph memory, and skill book memory. A variety of memory partitions and intelligent scheduling allow the AI to achieve a memory system that closely resembles that of humans.
Product DEMO
Chat with NPCs using stage mode
We utilized the stage's configuration mode to customize the dialogue logic for car sales. This demo demonstrates the feature's effectiveness in a specific scenario.
Chat with characters in fiction
This demo shows the effectiveness of talking with the characters in the fiction. It is a good practice of using skill book.
Voice communication with NPCs
We have voice to voice functionality while communicating with NPCs. The NPCs can send either voice or text depending on users' responses.