Introduction
Project AIRI
Presence without custody.
A private, persistent character who can remember, perceive, speak, appear, and act—without surrendering ownership of their life to a platform.
Built for AI companions, virtual characters, and anyone who wants more than a chatbot wearing an avatar.
A character, not a collection of features
AIRI is a character runtime.
The character has an identity that can be carried, senses that can be changed, a voice that can be replaced, bodies that can be exchanged, memories that can deepen, and tools that can extend what they are able to do. Those faculties may evolve independently. The character connecting them remains continuous.
One enduring presence across conversations, models, bodies, applications, and machines—owned by the person who shares a life with it.
The Character Stack
| Faculty | What AIRI Provides |
|---|---|
| Identity | Portable AIRI Cards, character profiles, acting cues, and per-character configurations |
| Continuity | 8-layer memory architecture: Short-Term Awareness (STMM), Text Journals, Lifetime Archives, and Echo Chips |
| Perception | Hearing, screen vision perception, image understanding, desktop awareness, and attention gating |
| Mind | Interchangeable cloud or local LLM brains, prompt synthesis, and cognitive streaming pipelines |
| Voice | Swappable speech (TTS) and transcription (STT) engines, voice profiles, Audio Studio, and sub-second live audio |
| Body | Live2D, VRM (3D), MMD, and Spine renderers with tactile touch, expressions, and wardrobe customization |
| Agency | Proactivity heartbeats, MCP tools, generative artistry, roleplay directors, and autonomous background loops |
| World | Desktop stages, docked Control Strip, web and pocket mobile companions, and Discord bot integration |
| Sovereignty | Local-first storage, portable character data, optional user-owned cloud sync (BYOS), and zero third-party telemetry |
Getting Started
AIRI supports both Web and Desktop platforms.
- Web Stage: Lightweight, instant access from any modern browser or mobile device without installation.
- Desktop Application: The full flagship experience with the docked Control Strip, local WebGPU/WASM inference (local LLMs, Whisper, Kokoro), background vision, and system sensor telemetry.
Contributing
For guides that help you understand how to contribute to this project, please refer to the Contributing page.
For references to help you design and improve the UI of Project AIRI, please refer to the Design Guidelines page.