Open LLM VTuber is unclaimed —

Open LLM VTuber

Real-time voice interaction with your own Live2D AI companion·docs.llmvtuber.com

Open source

Open LLM VTuber is a voice-interactive AI companion featuring a Live2D avatar (Cubism 5 support) that can hold real-time voice and text conversations, perceive its environment via camera and screen, and call external tools through the MCP protocol. It can run fully offline on local hardware or use cloud APIs, and supports Windows, macOS, and Linux through both a web client and a desktop client with a transparent desktop-pet mode. It integrates a wide range of LLM, ASR, and TTS backends and includes a Letta-based long-term memory system.

What it's for

Acting as a virtual girlfriend, boyfriend, or personal AI assistantReal-time voice and text conversation with an AI companionLive streaming interaction with viewers via a built-in BiliBili clientDesktop companion / pet mode that stays on screenCamera and screen-sharing based visual interactionBrowser automation controlled by the AIGroup chat with multiple AI characters

Features 20

AI-controlled browser automation with live view
AI-initiated proactive speech
Broad ASR backend support including sherpa-onnx, FunASR, Faster-Whisper, Whisper, Azure ASR
Broad LLM backend support including Ollama, OpenAI, Gemini, Claude, Mistral, DeepSeek
Broad TTS backend support including MeloTTS, Coqui-TTS, GPTSoVITS, Edge TTS, Fish Audio, Azure TTS
Built-in BiliBili live-stream client for danmaku interaction
Camera, screen recording, and screenshot-based visual perception
Cross-platform support for Windows, macOS, and Linux
Custom Agent interface for integrating architectures like HumeAI EVI, OpenAI Her, or Mem0
Desktop pet mode with transparent background, always-on-top, and click-through
Fully offline mode using local models
Group chat supporting multiple AI characters
Live2D interaction (Cubism 5) with clickable expressions
Long-term memory based on Letta
Persistent chat history across sessions
Tool calling via MCP protocol with real-time status display
TTS translation support (e.g., chatting in Chinese while AI speaks Japanese)
Visible AI 'inner monologue' (thoughts/expressions not spoken aloud)
Voice interruption without headphones
Web client and desktop client with switchable window/desktop-pet modes

At a glance

open sourceself-hostableAPI

TypeDesktop app

DeploymentHybrid

PlatformsWeb, Windows, macOS, Linux

LanguagesChinese, English

Integrations

SiliconFlowTTSGroq WhisperWhisper.cpppyttsx3SparkTTSGPTSoVITSOpenAI TTSOllamaHumeAI EVIMistralLettaFish AudioBarksherpa-onnxOpenAIAzure ASREdge TTSDeepSeekAzure TTSCoqui-TTSWhisperBiliBiliCosyVoiceLM StudiovLLMMeloTTSFunASRClaudeGeminiFaster-WhisperMem0

Resources

Docs