shinshin86/aituber-onair

Open source TypeScript toolkit for building Neuro-sama-style AI VTubers and AI characters that chat, speak, remember, and react to live viewers.

What it solves

AITuber OnAir provides a comprehensive toolkit for developers to build AI-powered VTubers (Virtual YouTubers) that can interact with live audiences. It solves the complexity of integrating LLMs, text-to-speech (TTS), avatar rendering, and viewer relationship management into a single, streaming-ready application.

How it works

The project is built as a modular TypeScript framework. You can start with a pre-built web app, use a CLI to scaffold a new project, or integrate specific packages into your own stack. It handles the full pipeline: receiving viewer comments, processing them through an LLM, generating emotional voice responses, and animating various avatar types (such as VRM, Live2D, or PNG) in real-time.

Who it’s for

This toolkit is for developers who want to create interactive AI characters for streaming platforms like YouTube or Twitch. It is designed for those who want to choose their own level of abstraction, ranging from quick-start templates to granular control over chat, voice, and agent logic.

Highlights

  • Supports diverse avatar formats including VRM, Live2D, Inochi2D, PNGTuber, and Mesh avatars.
  • Includes specialized modules for conversation quality, such as detecting repetitive patterns and rewriting responses to be more engaging.
  • Features a built-in relationship system to track viewer points, levels, and achievements.
  • Provides unified access to major LLM and TTS providers like OpenAI, Claude, Gemini, VOICEVOX, and VoicePeak.
  • Offers a CLI tool to scaffold projects with various templates in minutes.

Related

  • Project
    ronisarkarexe/story-spark-ai

    An open-source AI storytelling platform that generates multiple story variations from a single prompt and provides tools for analysis and comparison.

  • Project
    datawhalechina/leegenai-tutorial

    A comprehensive, beginner-friendly tutorial series based on Professor Hung-yi Lee's university courses that covers the fundamentals, architecture, and practical applications of generative artificial intelligence.

  • Project
    AnswerDotAI/claudette

    A high-level Python wrapper for Anthropic's SDK that simplifies stateful chat, tool use, and prompt caching for Claude models.

  • Project
    IBM/prompt-declaration-language

    A declarative YAML-based language for creating reliable, composable LLM prompts and integrating them into software systems with support for control flow and code execution.