XiaoGPT lets you talk to ChatGPT and other large language models through a Xiaomi AI Speaker, turning the smart speaker into a voice front end for modern LLMs. Ask a question out loud and the answer is streamed back in speech — no phone or keyboard involved. The project has about 6,900 stars on GitHub and is actively developed in 2026.
The Python service connects to your Xiaomi account and speaker, listens for questions routed through the speaker, sends them to the LLM backend you configure — ChatGPT and several other models are supported — and speaks the reply using text-to-speech, including options such as Edge TTS and Azure voices. Responses stream token by token so the speaker starts answering quickly, wake-up and continuous-conversation modes let you keep a dialogue going hands-free, and prompts and personas can be customized so the speaker answers in a chosen style or role. It also supports extras such as playing replies for long-form content and integrating with the speaker’s native abilities. Setup runs on any always-on machine — a Raspberry Pi or a home server is typical — and configuration is done through environment variables and a small config file. XiaoGPT is MIT licensed and is a favorite home-automation hack in the Chinese open-source community.
XiaoGPT is free and open source under the MIT license. You need your own Xiaomi speaker and your own LLM API account.
XiaoGPT is for smart-home tinkerers who own a Xiaomi AI Speaker and want it powered by a real LLM — for questions, conversation and fun at home — instead of its limited built-in assistant.