Using security camera audio feed for assist and wake word detection

Configuration Voice Assistant
1

I already have a frigate nvr and cameras in my house, as well as smart speakers in most rooms. Has anyone managed to split the audio stream from rtmp rstp feeds and use them as mics for assist? Furthermore using arbitrary media players for audio feedback?

2

I figured it out, ill just use an existing docker container i made with pulseaudio to pipe an ffmpeg stream into a virtual microphone. Then use the homeassistant satellite to listen in. Rince and repeat for each camera in the house.

3

Hello

Could you give us more détails ?

i use StreamAssist for now: AlexxIT/StreamAssist: Home Assistant custom component that allows you to turn almost any camera and almost any speaker into a local voice assistant (github.com)

2 Likes
4

Ill play around with it. But i want the mic to use wake word detection

I have a docker container that runs a music streaming client which plays to a dummy pulse audio sink, which then is streamed to icecast with ffmpeg. I then tell sonos to play the flac stream for audio output.
Theres one container per room/speaker

So my idea is to add snapcast and a assist satellite to the docker container and stream the audio channel to a dummy source so its always listening for wake word dectection.