Open-LLM-VTuber/open-llm-vtuber ? reverse-engineered prompt

Build me a local AI companion app that I can talk to with my voice and hear back in real time.

It should work on Windows, Mac, and Linux, and have both a web view and a desktop app. I want a Live2D character on screen that can show expressions, react while I talk, and also work in a small pet mode with a transparent background so it can sit on my desktop. Make sure it supports voice interruption, so I can cut it off naturally while it is speaking, and let it keep chat history so I can come back to old conversations later.

I want it to run fully offline if I choose, using local models, but also be able to connect to common AI, speech to text, and text to speech services if needed. Please make the character and persona easy to customize, and include a simple setup flow with sensible defaults. If you need to check current docs for model setup or platform quirks, go ahead and look them up online.