gpt-home

Turns a Raspberry Pi into a voice assistant powered by ChatGPT. Plug in a microphone and speaker, run the install script, and you've got a smart speaker that uses OpenAI's API instead of Alexa or Google Home. Say something, it sends it to GPT, speaks the response back.

The setup is a Docker container on a Raspberry Pi (3B+ or 4 recommended). It handles wake word detection, speech-to-text, GPT completion, and text-to-speech in a pipeline. Supports Home Assistant integration so you can control smart home devices by voice.

Completely free software. Your cost is hardware (a Pi you probably already have) and OpenAI API usage. At casual home use (maybe 20-30 queries a day), expect $1-3/mo in API costs.

The catch: this is a hobby project. It works, but don't expect Alexa-level reliability or polish. Wake word detection isn't perfect. Latency is noticeable; you're making a round trip to OpenAI's API for every query. There's no offline fallback. And if OpenAI changes their API or pricing, you're dependent on the maintainer updating. For a more robust self-hosted voice assistant, Home Assistant's own voice pipeline is more mature.