VirtualWife is an open-source project aiming to create a virtual digital human with a "soul," designed for companionship, emotional support, and integration into daily life. It targets users seeking AI-driven interaction, offering features like Bilibili live streaming, LLM integration (OpenAI, Ollama), and VRM model support.

How It Works

The project leverages a modular architecture to enable customizable virtual companions. It supports text-to-emotion and text-to-action driven by LLMs, with options for various speech synthesis engines (Microsoft Edge, Bert-VITS2) and a focus on low-latency, stream-based data processing for responsive interactions. Long-term and short-term memory modules are integrated to enhance conversational continuity and AI personality.

Quick Start & Requirements

• Installation: Docker-based deployment is the primary method.
  • Docker installation guide
  • Docker Compose installation guide
• Prerequisites: Docker and Docker Compose.
• Setup: Initial startup may take up to 5 minutes for image downloads.
• Access: Web interface available at http://localhost/.

Highlighted Details

• Supports Bilibili live streaming integration.
• Allows switching between multiple LLM providers, including private Ollama models.
• Features character customization with VRM model support from Vroid.
• Includes text-driven facial expressions and body movements.

Maintenance & Community

The project is actively developed, with a roadmap indicating planned improvements for memory optimization, emotional expression, and character role-playing. Community interaction is facilitated via a technical exchange group.

Licensing & Compatibility

Licensed under the MIT license, allowing for broad use and modification with the caveat that users assume all risks. This license is generally permissive for commercial use and integration into closed-source projects.

Limitations & Caveats

The project is described as being in the "incubation stage" with many areas for optimization. Specific features like advanced memory recall, personality shaping via LoRA, and a dedicated reflection module are still under development.