Core Technology
Core Technology
We design memory and conversation flows that help AI feel at home in a service.
Hierarchical Memory Structure
Human-like memory system with separated short, mid, and long-term memory
sLLM/LLM Dynamic Routing
Optimal model selection based on context for cost efficiency and quality
State Tag-Based Model Selection
Dynamic response model selection based on emotion and knowledge states
Selective Forgetting & Sleep Reconstruction
Memory refinement based on importance and sleep-phase reconstruction
LoRA-Based Character Identity
Consistent character personality and style through multimodal LoRA
Multi-Agent Autonomous Interaction
Natural conversations and interactions between characters
User-Based Memory Isolation
Independent management of personalized relationships and memories per user
Where the technology actually runs
Patents and core tech aren't slides — they're proven in products running right now. See which service each capability powers.
Conversational LLM · AI characters
Natural conversation with AI characters that grasp emotion and context — the core interface across several services.
Dynamic multi-model routing
Dynamically picks from 15+ LLMs (OpenAI·Claude·Gemini) per request to balance quality and cost.
Hierarchical long-term memory
A hierarchical memory structure that keeps characters in context across long chats and writing. A core patent-pending technology.
Generative AI (image·text)
We run image- and text-generation pipelines reliably in production.
On-device · Computer vision
On-device AI and vision (OCR, photo→floorplan, image processing) protect privacy and speed.
Global multi-language
Designed multilingual (KR·JP·EN) from the start to launch globally right away.
IbouAI has provisionally filed 9 patents for its core technology.