Hi i want to build up a voice control like Alexa but with data-privacy in order to control covers etc via homeassistant .
I know that it will not be that nice and performant like Alexa and Goole but then i have a data-privacy.
My setup is a own PI4 with respeaker 4 mic and a external speaker.
Now i am searching what will be the best voice-contol API.
I have focused now on:
- Almond & Ada (seems to be complex but powerfull?)
- rhasspy
- mycroft (picroft) seams to be easy but very slow?
Who has experience in this project and which API would you recommend?
Thanks
Ingo
Hey
i am currently using rhasspy
i do not know others but as far as i can say : it works pretty well : recognition is fine, wake world is customisable
i tested it in english and then in french
1 Like
Hi Ingo, sorry to be so late to the party here.
I also use Rhasspy and Home Assistant, and do recommend the combination. Quality isn’t as good as Alexa or Google devices (of course not with their resources), but it all runs locally. Background noise is a factor with all devices.
I did find that HAOS uses Docker to create virtual machines for HA and Rhasspy - but the magic needed to give a VM access to the hardware was beyond my reach at the time. Instead I went with running the compute-intensive modules on the RasPi4 as a base station (no mic or speaker directly connected); and each satellite RasPi is basically just for microphone and speaker. The same Rhasspy app runs on all devices, but with different modules enabled/disabled.
I did find it a little confusing to get setup properly, but all the requred info was in the documentaton (even if it took me three reads before it fell into place) - but that seems to be a feature of FOSS 
The Rhasspy forum has gone quiet lately, but announced today the primary Rhasspy developer will start working for Nabu Casa (the commercial arm of Home Assistant), so we are expecting development to pick up, with a “better out-of-box experience” being a goal.
Cheers,
Don
1 Like