CyberVerse
Self hosted, real-time digital human agent platform. Build voice-first AI agents with WebRTC, persona memory, tools, RAG, and optional digital-human video.
What is CyberVerse?
Self hosted, real-time digital human agent platform. Build voice-first AI agents with WebRTC, persona memory, tools, RAG, and optional digital-human video.
Overview of CyberVerse
What it is
CyberVerse is an open-source, self-hosted real-time digital-human agent platform. It enables developers to build voice-first AI agents using WebRTC for low-latency communication, persona memory, tools, retrieval-augmented generation (RAG), and optional digital-human video capabilities. The platform is designed to create interactive, conversational AI characters that can see, hear, and speak in real time, similar to a video call with a real person.
Key features
- Real-time digital-human video interaction: With a single photo, users can create a digital human ready for full-duplex video conversation. The platform integrates local models (FlashHead, LiveAct) and cloud services (Baidu Xiling, Xunfei Digital Human, Vidu S1).
- PersonaAgent + SubAgent tasks: A multi-agent architecture where PersonaAgent handles foreground conversation and interruptions, while background SubAgents asynchronously perform long-running tasks such as search, research, summarization, and HTML report generation.
- Character memory and RAG: Conversation history is persisted to local disk and automatically loaded across sessions. Users can import knowledge bases, documents, and biographical material for retrieval-augmented generation, aligning answers with the character's persona.
- Plugin-based stack: Brain, voice, hearing, tools, memory, and face are replaceable modules. Runtime behavior is configured in
config/cyberverse.yaml, while provider definitions are loaded from built-in directories and optional local overrides. The LiteLLM plugin provides access to 100+ LLM providers through a single interface.
Use cases
CyberVerse is suitable for building interactive AI agents that require real-time voice and video interaction, such as virtual assistants, digital companions, customer service avatars, and character-driven applications. It can also be used for preserving or recreating the appearance and voice of a person from a single photo, enabling emotional or memorial applications. The platform supports both local and cloud-based digital-human models, making it adaptable for various deployment scenarios.
Getting started
See the repository README for detailed installation steps. The prerequisites include Node 18+, Go 1.25, Conda, Python 3.10+, FFmpeg, and several system libraries. For a quick start, users can launch CyberVerse from cloud images provided by Compshare or AutoDL. The basic local setup involves cloning the repository, creating a Conda environment, copying configuration files, and setting environment variables for API keys.
License and pricing
CyberVerse is licensed under GPL-3.0. It is open-source and free to use under the terms of that license. Pricing for cloud services (e.g., digital-human APIs, GPU rental) is determined by the respective providers and is not included in the CyberVerse project itself.
Based on the project's public README and repository metadata.