Meet AIRI — Your Anime AI Girlfriend Anywhere, Anytim
AIRI lets anyone design a unique AI companion, bring it to life as a lovable desktop character, and invite it to join conversations on Discord, Telegram, Minecraft, and more — all while keeping full control of data and privacy.
- Always With You: AIRI lives on your desktop and favourite chat platforms, ready to talk, play, and help whenever you are online.
- Fully Yours: Shape personality, voice, and appearance in minutes. No coding required.
- Privacy First: Speech recognition and voice activity detection run locally, so private moments stay on your device.
- Open & Extensible: AIRI is open-source and modular — builders can add new skills or connect to additional AI models with ease.


Project Information
Field | Details |
---|---|
GitHub Repository | 🔗 moeru-ai/airi |
Live Demo | 🌐 airi.moeru.ai |
Author | @proj_airi |
Project Status | ⭐ 14.3k Stars · 🍴 1.2k Forks · 🔄 Active |
License | MIT License |
Core Experience in 30 Seconds
- Launch the App → A transparent window appears; your AI avatar looks up and blinks.
- Say “Hi AIRI” → The app detects your voice, transcribes it locally, and the AI answers back in real-time speech.
- Personalize → Open Settings to tweak the character card, swap a 3D/Live2D model, or choose a different AI provider.
- Extend → Type
/summon
in Discord or invite the Telegram bot — your companion joins the conversation as a natural participant.
At a Glance: What You Can Do
You want to… | AIRI makes it possible |
---|---|
Have a chatty desktop companion | Real-time voice & text conversation with gaze tracking and click-through window so it never gets in the way |
Craft a unique persona | Character Card editor lets you define backstory, tone, and catch-phrases |
Control every detail | Choose local or cloud LLM/TTS models, upload custom avatars, pick fonts and colours |
Bring the AI to social spaces | One-click bots for Discord voice, Telegram groups, and even Minecraft servers |
Experiment & build | Prompt-engineering playground and modular architecture for developers |
How It Works (Conceptual)
- Voice Flow – Your microphone feeds a lightweight on-device model that decides when you are talking. Captured audio is transcribed locally, then sent to the AI brain (LLM). The response streams back and is spoken aloud through a TTS service.
- Desktop Magic – The character is rendered in a frameless window. When the mouse isn’t hovering over the avatar, the window becomes “click-through,” so you can keep working normally.
- Personality Engine – A simple “Character Card” file stores traits, speech style, and example dialogue. Every AI request combines this card with recent chat history, keeping the character consistently in-role.
- Platform Bridges – Stand-alone bot services translate platform events (Discord messages, Telegram updates, Minecraft world state) into a common format, letting the same AI brain respond everywhere.
Takeaway: All the complexity (audio pipelines, window hacks, multichannel bots) happens behind the curtain. As a user you only see a friendly face that remembers you and shows up wherever you hang out online.
Next Steps
- Getting Started: Follow our 5-minute setup guide to connect a speech model and choose an avatar.
- Desktop Companion Experience: Discover how AIRI creates an unobtrusive desktop presence.
- Understanding AIRI’s AI Brain: Learn how AIRI maintains personality and consistency.
- Platform Integrations: Extend AIRI to Discord, Telegram, and Minecraft.
- Developer Guide: Technical deep-dive for builders and contributors.
Related Projects on HowWorks
- AI Town: Explore a16z’s living AI character simulation
- AI Hedge Fund: Build AI trading strategies without coding
- Onlook: The visual editor that understands your design intent