Repolode
Submit
moeru-ai

moeru-ai/airi

Self-hosted, self-owned AI companion (VTuber-style virtual character) with realtime voice chat, game play, and Live2D/VRM avatars for Web, macOS and Windows.

Open-source alternative to:Character.ai

View on GitHubWebsite
Stars
50,538
Forks
5,078
Open issues
239
Last commit
5 hours ago

What is it?

AIRI is an open source attempt to recreate Neuro-sama, an AI virtual streamer, as a personal companion you run yourself. It combines a character that can talk, listen and be displayed as a Live2D or VRM avatar, with the ability to chat in Discord and Telegram and to play games such as Minecraft and Factorio.

The project offers several front ends: a browser version, a desktop app for macOS and Windows built on Electron, and an experimental mobile version built with Capacitor. Speech input uses browser audio and client-side recognition, and speech output supports several TTS providers, including ElevenLabs, Azure Speech and local Kokoro TTS.

For the language model, AIRI connects to a wide list of LLM API providers such as OpenAI, Anthropic Claude, Google Gemini, DeepSeek, OpenRouter, Ollama and vLLM. It is built in TypeScript as a pnpm monorepo, with the code and documentation under the MIT license.

Key features

  • 01Realtime voice chat with browser audio input and client-side speech recognition.
  • 02Live2D and VRM avatar control with auto blink, auto look-at and idle eye movement.
  • 03Multi-provider text-to-speech, including ElevenLabs, Azure Speech and local Kokoro TTS.
  • 04Connects to many LLM providers such as OpenAI, Claude, Gemini, DeepSeek and Ollama.
  • 05Chat integrations with Discord and Telegram.
  • 06Game play demos for Minecraft and Factorio.

Use cases

  • Building a self-hosted virtual companion or digital pet that runs on your own machine.
  • Experimenting with AI VTuber or streamer characters using Live2D or VRM avatars.
  • Prototyping a voice-driven chatbot that speaks through several TTS providers.
  • Studying how a web-based app combines LLMs, speech and 3D or 2D avatars.

Quick start

  1. 1

    Install on Windows with winget

    winget install MoeruAI.AIRI
  2. 2

    Install on macOS with Homebrew Cask

    brew install --cask airi
  3. 3

    Or download the installer from the releases page

  4. 4

    Or try it in the browser at airi.moeru.ai

Things to know

  • The project is in an early stage of development, so expect rough edges and changes.

Topics

#ai-companion#vtuber#live2d#vrm#llm#text-to-speech#typescript#electron#ai-vtuber#airi

FAQ

What is airi?
Self-hosted, self-owned AI companion (VTuber-style virtual character) with realtime voice chat, game play, and Live2D/VRM avatars for Web, macOS and Windows. AIRI is an open source attempt to recreate Neuro-sama, an AI virtual streamer, as a personal companion you run yourself. It combines a character that can talk, listen and be displayed as a Live2D or VRM avatar, with the ability to chat in Discord and Telegram and to play games such as Minecraft and Factorio.
Is airi free to use? What license does it have?
airi is open source under the MIT license. Check the license terms before using it in commercial products.
Is airi still maintained?
Actively maintained. The last commit to airi was on Oct 21, 2026.
What can I use airi for?
Building a self-hosted virtual companion or digital pet that runs on your own machine. Experimenting with AI VTuber or streamer characters using Live2D or VRM avatars. Prototyping a voice-driven chatbot that speaks through several TTS providers. Studying how a web-based app combines LLMs, speech and 3D or 2D avatars.
How do I install airi?
Install on Windows with winget: winget install MoeruAI.AIRI Install on macOS with Homebrew Cask: brew install --cask airi Or download the installer from the releases page Or try it in the browser at airi.moeru.ai
Where can I get airi?
The source code and installation instructions are on GitHub: https://github.com/moeru-ai/airi

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects