Airi: Open-Source AI VTuber for Real-Time Interaction

Airi: Bringing AI Virtual Characters to Life in Your Browser and Beyond

Airi is an innovative open-source project dedicated to creating advanced AI waifus and virtual characters designed for real-time interaction, including engaging in voice chats and even playing popular games like Minecraft and Factorio. Inspired by projects like Neuro-sama, Airi aims to achieve a new altitude in AI-driven virtual personalities.

Built for Accessibility and Performance

What sets Airi apart is its foundational reliance on modern web technologies such as WebGPU, WebAudio, Web Workers, WebAssembly, and WebSocket. This robust web-centric architecture ensures that Airi can run smoothly not only on desktops but also directly within modern browsers and even on mobile devices with Progressive Web App (PWA) support. This broad compatibility opens up numerous possibilities for developers and users alike, offering flexibility without sacrificing advanced features.

Current Capabilities and Future Vision

Despite being in an early stage of development, Airi already demonstrates impressive capabilities:

• Brain: Integrated with various LLM API providers like OpenAI, Google Gemini, Anthropic Claude, and more, allowing for intelligent and dynamic conversations. It also boasts the ability to play Minecraft and Factorio, and chat on platforms like Telegram and Discord.
• Memory: Features pure in-browser database support (DuckDB WASM, pglite) and pure in-browser local (WebGPU) inference, laying the groundwork for sophisticated memory recall and learning.
• Ears & Mouth: Equipped with client-side speech recognition for audio input from browsers and Discord, as well as ElevenLabs voice synthesis for natural-sounding output.
• Body: Supports both VRM and Live2D models, offering detailed control over virtual character animations, including auto-blinking, eye movement, and looking at interactions.

How to Get Involved

The Airi project is a collaborative endeavor actively seeking talented individuals to contribute. Whether you're a seasoned developer proficient in Vue.js and TypeScript, an artist skilled in Live2D or VRM modeling, a VRChat avatar designer, or an expert in computer vision, reinforcement learning, speech recognition, or synthesis, there's a place for you. The project encourages experimentation and contribution, even for those more familiar with other frameworks like React or Svelte.

By joining the Airi community, you can help shape the future of AI companions, pushing the boundaries of what's possible in interactive virtual experiences. Explore their GitHub repository to learn more about development, current progress, and how to contribute to this exciting open-source initiative.