VirtualWife

Why we included this project

VirtualWife bundles most of the plumbing for running an AI companion with a face and a voice. It ships as a set of Docker containers that combine a large language model backend with a VRM 3D avatar, and you can point it at OpenAI or run everything locally through Ollama. On top of plain chat it has short- and long-term memory, text-driven expressions and gestures, and switchable Chinese speech synthesis via Microsoft Edge or Bert-VITS2. The standout feature is the Bilibili livestream hookup: creators who want an AI VTuber-style persona on stream get a working reference path instead of having to assemble the pieces themselves. Keep expectations realistic. The author describes it as early-stage, so treat it as a foundation to build on rather than a finished product.