9 Provisional Applications

Patent-Pending Technology

IbouAI has provisionally filed 9 patents to protect the core technology behind its services.
We are securing core IP for next-generation AI character services, including hierarchical memory, state-tag routing, NFT-based character execution, and autonomous multi-agent interaction.

Patent #01 Model Routing

Method and system for providing a multimodal AI agent service using state-aware lightweight language model routing and a hierarchical memory structure

State-Aware Lightweight Language Model Routing and Hierarchical Memory Structure
Status: Provisional Application

Optimizes operating costs by using sLLMs and LLMs together, and generates personalized multimodal content (text, images, video) by tiering user state information into short-, medium-, and long-term memory.

Key Elements
  • ▸ Query-complexity-analysis routing
  • ▸ Hierarchical memory (short/medium/long-term)
  • ▸ Multimodal integration engine
BestCharAljooB2B SI
Patent #02 State Tag Control

Method and system for selecting an advanced language model and controlling multimodal output using a language model that generates state tags

Advanced LLM Selection and Multimodal Output Control via State Tags
Status: Provisional Application

The language model generates emotion/knowledge-exploration state tags as inference results, and uses them to control whether an advanced language model is invoked and the multimodal output method (image tone, voice style).

Key Elements
  • ▸ Generation of emotion/knowledge state tags
  • ▸ Dynamic invocation of the advanced model
  • ▸ Synchronization of visual/auditory style
BestCharFictionDay
Patent #03 Voice Agent

Character-selectable virtual agent system that performs autonomous voice interaction based on relationship memory and emotional/behavioral state

Character-Selectable Virtual Agent for Autonomous Voice Interaction
Status: Provisional Application

The user selects a virtual character to initiate voice interaction, voice expression is dynamically controlled according to per-user relationship memory and emotional/behavioral state, and under certain conditions the character proactively initiates conversation.

Key Elements
  • ▸ Character-selectable voice connection
  • ▸ Per-user isolation of relationship memory
  • ▸ Initiation of autonomous interaction
AljooSmart SpeakerB2B
Patent #04 Blockchain / NFT

Method and system for AI character execution and relationship management using tokenized-digital-asset-based execution authorization and relationship-state anchoring

NFT-based Execution Authorization and Anchoring of Relationship State
Status: Provisional Application

Verifies AI character execution authorization with tokenized digital assets such as NFTs, and decrypts an encrypted character package with a one-time/time-limited execution key. Anchors a relationship-state summary hash to a distributed ledger to provide integrity and timestamp proof.

Key Elements
  • ▸ Token-based authorization verification
  • ▸ Encrypted character execution package
  • ▸ On-chain anchoring of relationship state
  • ▸ Memory-handling policy upon transfer
BestChar IPNFT Marketplace
Patent #05 Memory Management

Human-like AI agent system with a selective forgetting algorithm and sleep-phase-based memory re-encoding for optimizing user engagement

Human-like AI Agent with Selective Forgetting and Sleep-phase Memory Re-encoding
Status: Provisional Application

Mimics human cognitive structure to selectively forget memories based on importance, usage frequency, and elapsed time, and re-encodes memory during the agent's idle time (sleep phase), simultaneously achieving a natural conversational flow and reduced operating costs.

Key Elements
  • ▸ Selective forgetting algorithm
  • ▸ Sleep-phase-based memory re-encoding
  • ▸ Retroactive memory recall
BestCharAljooFictionDay
Patent #06 Multi-Agent

Autonomous, reactive multi-agent interaction and relationship-memory management system based on the exchange of emotional-state and action tags

Autonomous Multi-Agent Interaction via Emotion and Action Tag Exchange
Status: Provisional Application

Multiple AI agents generate and exchange emotional-state and action tags to autonomously determine reactions without central control, isolate and manage per-user relationship memory, and implement a natural multi-party conversational rhythm through continuous-utterance control.

Key Elements
  • ▸ Autonomous reaction without central control
  • ▸ Mutual recognition based on tag exchange
  • ▸ Continuous-utterance control
FictionDayGames/MediaB2B Content
Patent #07 Multimodal Generation

Content generation method and system that dynamically controls multimodal parameters for character representation based on state tags

Dynamic Multimodal Character Representation via State Tags (LoRA)
Status: Provisional Application

While maintaining reference-image-based character identity, selectively and dynamically controls image/video/voice generation parameters or low-rank adaptation modules (LoRA) according to state tags generated by the language model.

Key Elements
  • ▸ Maintaining character identity
  • ▸ Dynamic LoRA swapping
  • ▸ Multimodal synchronization
BestCharAljoo
Patent #08 Adaptive Agency

Human-like AI character system in which the level of proactive agency changes according to memory-isolation-based relationship formation

Memory Isolation-based Relationship Formation and Adaptive Agency
Status: Provisional Application

While sharing the same character knowledge structure, isolates memory by user identifier to form distinct relationships per user, and the level of proactive agency changes dynamically according to the degree of relationship accumulation.

Key Elements
  • ▸ Per-user memory isolation
  • ▸ Formation of relationship state
  • ▸ Adjustment of proactive agency level
BestCharAljoo
Patent #09 NFT / Interop

Method and system for one-click multimodal AI character generation and provision using an NFT-based character execution package

One-Click Multimodal AI Character Generation via NFT-based Execution Package
Status: Provisional Application

Registers character images, prompts, and execution metadata on an NFT, and with one click automatically composes and generates multimodal character output including chat, emotion, images, video, and voice. Ensures interoperability across services via SDK/API.

Key Elements
  • ▸ Character execution package standard
  • ▸ One-click multimodal execution
  • ▸ Cross-service interoperability
BestCharNFT MarketplaceExternal Platforms

Notice

All of the above patents have been provisionally filed with the Korean Intellectual Property Office, and the official application numbers will be updated sequentially as the examination process proceeds. The descriptions are written within the scope of the published specification summaries, and the specific claims may change.

Are you interested in B2B licensing or SI collaboration using our patent-pending technology?

Contact Us →