tegnike/aituber-kit
Chat, Stream, or Kiosk: A Universal Remote for AI Characters
This toolkit wires up 15 LLM and 11 voice APIs to VRM, Live2D, or PNGTuber avatars so you can deploy a chat, stream, or kiosk app without writing the plumbing.
Collecting fresh signals — velocity needs a few days of history.
collecting data…
star history
What it does
AITuberKit is essentially a highly configurable Next.js-based integration layer. It provides a web interface where you pick your LLM backend, your text-to-speech engine, and your avatar format—VRM, Live2D, or video-based PNGTuber—and get a talking AI character that can hold conversations, read YouTube live comments, or run as a fullscreen digital signage terminal with face-detection greetings. The README calls it an “all-in-one toolkit,” which largely means it handles the protocol negotiation and state management between services you already have API keys for.
The interesting bit
The project treats “AI character” as a systems-integration problem rather than a machine-learning one. The standout modes are the practical ones: a game-commentary mode that captures your screen and has the avatar riff on it, and a digital-signage mode with motion-sensor-style face detection to trigger welcome and goodbye phrases. It is less about inventing new AI and more about inventing new contexts for existing AI.
Key highlights
- Supports 15 LLM providers (OpenAI, Anthropic, Google Gemini, Groq, local options via Ollama/LM Studio, etc.) and 11 TTS engines (VOICEVOX, ElevenLabs, Style-Bert-VITS2, etc.).
- Avatar formats include 3D VRM with motion tags, Live2D (Cubism 3+), and MotionPNGTuber video sprites.
- Specialized modes for YouTube streaming comment response, slide presentation narration, WebSocket external integration, and screen-capture game commentary.
- Deployment targets include Vercel, Cloudflare Workers (via OpenNext), and Docker, though the README notes a custom license applies for commercial use since v2.0.0.
Caveats
- The README carries a prominent security warning that the repository is intended for personal or properly secured commercial use; running it as-is on the open web appears to be discouraged without hardening.
- Since v2.0.0, the project uses a custom license, so commercial deployment requires reading the specific terms rather than assuming standard open-source freedoms.
Verdict
Ideal for developers who want to prototype AI hosts, streamers, or interactive kiosks without building yet another chat frontend. Skip it if you are looking for novel model training or a fully managed SaaS; this is a self-hosted wiring harness.