I’m new to this, so apologies for any dumb questions or comments.
I want to use my ReSpeaker 4-mic array (connected to a Rpi Zero 2W) to listen to my voice commands, but I want the assistant verbal response to come through my Sonos speakers. I’ve been working on this for about 2 weeks (using mostly ChatGPT for guidance), have sunk about 30 hours into this, and am beginning to wonder if this is even possible?
Thanks
That might be your problem. HomeAutomation area is so fast moving most AI slop is outdated and often wrong. You have to go back, read the real documentation, configure it carefully, and understand what you are doing. Trying to battle with AI slop is often an added amount of pain you don’t need.
Having said that, there are no dumb questions. You may just need to keep throwing them, gradually closing in, till you hit the target.
Let’s think of this in a logical manner, no artificial involved.
You have a number of devices listed. Is each working well individually? Does HomeAssistant recognise them all?
What are your entities named?
How are you getting input into HomeAssistant? What devices. What integration?
How are you processing it?
How are you getting it out again?
Getting each part going is one challenge. Getting them working together is a higher one.
Does turning on debugging and looking at logs offer any clues?
Tell us what you have done so far, and what your current bottleneck is.
Thanks for the response and recommendations. Agreed: While I have been able to get many things “happening” by using AI, it’s not “coming together” because I really don’t understand it like I need to. It was an initial, lazy approach to getting a result. Because I’ve spent about 30 hours now on this, I do know much more than I did initially. Will take the time to read through all the documentation–HA seems to be extremely powerful and configurable and I hope to use it long-term. Thanks again.
PS No way to edit my initial post is there? I misspelled “Commands” and want to fix it.
Also, let me ask this: Has ANYONE used a ReSpeaker (mine’s connected to a Rpi-Zero 2W) as the assist microphone and routed the HA assist response to another speaker (e.g., a Sonos)? If it’s currently impossible, it will save me many more hours because I’m kind of on a mission to make this work.
You may find the Zero is severely constrained by both memory and processor speed issues for your needs.
Yes, I’m using it as a “starting point” (I had tried an ESP32 + I2S mic initially) - I just want to get a working ReSpeaker Mic (input) and be able to route the Assistant Response to my Sonos speakers. I see people in the community have done this with the Voice PE (which I also have) but the mic (& speaker) on it is insufficient. I can’t find anyone who claims to have accomplished the ReSpeaker/Sonos combo that I’m trying to achieved.
Sonos won’t announce flac files. There’s an open issue on it.
I have two ReSpeakers which plays their responses on a Sonos speaker.
Look at FormatBCE’s work: GitHub - formatBCE/Respeaker-Lite-ESPHome-integration
He has a blueprint for HA that you can use choose which ReSpeaker outputs to which media_player.
This looks amazing - thank you! I’ll try to get it working then will let you know - thanks again
Not sure this is exactly what you are looking for but I set up a BeeLink S13 with HAOS and a XVF3800 ReSpeaker from Seeed. It works “ok” as a satellite system to my server HAOS hosting all the automations and such. I use REST to link the two, sending only text between them. It works fine, it’s just that the BeeLink takes about 2.6 seconds to process about 1.5 second command. So from wake word activation to text command sent to my server is about 5-6 seconds for ~1 second speech command. Using small-int8 model. I will likely start v2 using a more powerful system and incorporate Home-3B LLM into it in a few months. As a retired software engineer working with voice systems 25+ years ago, the obvious place to increase performance is not just with faster host but with concurrent processing. Current designed system components block on each step. Whisper should be getting an audio stream the moment I start talking, not a ms after VAD decides I stopped a half second ago.
Greetings,
I’ve now put about 40+ hours trying to get this to work and would appreciate any tips or advice you can give.
Other Notes:
ReSpeaker Lite appears in my Device Manager under USB port
See my ReSpeaker YAML (img 01)
See secrets.yaml (img 02)
See respeaker in my discovered integrations (img 03)