Whisper

A Home Assistant voice assistant needs three parts. Something to turn speech into text, something to turn a reply back into speech, and Assist in the middle to work out what you meant. The first two can run on plain CPU in Docker, so nothing you say has to leave the house. I built the whole stack from scratch on 11 October 2026 on Home Assistant 2026.10.1, ran text and spoken commands through it, and timed the result. This post has the compose file, the settings, the numbers and the two things that tripped me up. ...

Updated October 2026: I re-tested this on 8 October 2026 against the current rhasspy/wyoming-whisper image. The compose file no longer needs a version line, the default model is now small-int8 instead of medium-int8, and there’s a table of measured model speeds, a test script and a troubleshooting section. The Home Assistant steps come from my original post. I haven’t re-run them on a current Home Assistant release. Wyoming Whisper is the speech-to-text half of a fully local Home Assistant voice assistant. It wraps faster-whisper, a fast implementation of OpenAI’s Whisper models, and Home Assistant talks to it over the Wyoming protocol. The text-to-speech half is Wyoming Piper. ...