I'm sure it's a decent product, but only 30 days warranty?? They must not have a lot of faith in their product. That's not even legal in a lot of countries (at least the EU, Australia, New Zealand, and some Asian countries).
post
Very cool!
The voice sattelites like this and the Home Assistant voice from a year back are great and I love seeing more options for self hosted fully local devices like this.
But honestly, what I find is missing for a fully local smart home setup is a "brain" for running the assistant LLM in a simple way.
There is plenty of cloud solutions for connecting AI, but true local hosting requires a rig with a proper GPU if you want timely responses. I wonder if it is possible to build a purpose built "brain box" for Home Assistant that is small, not too expensive, fully local.
It just sucks needing a rig with a RTX 3090 or whatever to get the full chain to run fully local. Small LLMs are stupid as dirt and our Home Assistant Green is in no way equipped to handle any form of LLM. I managed to get my gaming rig to host an LLM that was good enough, but I don't want my gaming rig to be always on, and if I start gaming on it, the GPU has to deallocate all LLM business.
I love fully local stuff, but the LLM part seems very expensive. Even for such a simple thing as managing our lights and music.
I'm just gonna shamelessly plug my project here: https://github.com/charludo/hass-closest-intent
You do need local STT, but it's fine if it's somewhat bad; the conversation agent linked is sufficient to "get" what you meant, without LLM.
I really appreciate the clarity and straightforward intro in the readme.
I love fully local stuff, but the LLM part seems very expensive. Even for such a simple thing as managing our lights and music.
So, the simple things should just be voice command (not recognition, limited vocabulary), kitchen light on, what's todays weather (assuming you automatically download it and have it there, needs TTS like whisper). That can run on a recent potato. Somewhat more complex things need full voice recognition e.g. Play me x songname by y artist. Laptop CPU should handle that fine. You'll only need a full LLM for more general inquiries, e.g. What mess has Trump made today?, which will need web access seeing as small LLMs don't have a lot of world knowledge (although you can self host Wikipedia and point it at that).
If you have a desktop computer and a GPU with 16+GB VRAM (or even smaller using RAM and a MoE LLM) that'll do the job with reasonable smarts e.g. Qwen 3.6 35BA3B, Gemma 4 12B or 26BA4B. You may have to wait a minute or two. Mine pulls around 55W at idle (7800XT) and does double duty as a NAS and gaming rig, or you can have it sleep most of the time and let the laptop wake it up when necessary.
I actually have things like "play x by y" functioning really well without a LLM.
Have a custom service that exports all song/album/artist names from MusicAssistant, does some simple cleanup, and places the list where HomeAssistant expects it for custom voice intents. Then this: https://github.com/charludo/hass-closest-intent is enough that imperfect STT can still easily be matched to those song/artist/... names.
“how much is using local LLMs for trivial things raising world wide energy usage and causing a RAM shortage?”
Get a Strix Halo box to be your headless AI server
They're around 4k USD these days though...
I love fully local stuff, but the LLM part seems very expensive. Even for such a simple thing as managing our lights and music.
Yeah, this is something I found out when I hacked my Amazon Echo and put LineageOS on it. I like the new interface that isn't constantly advertising Amazon bullshit as a screensaver, but it doesn't have a GPU, so attempting to put any sort of voice component takes many seconds to try to process.
And in this post-memory-crisis economy, it's not quite the right time to buy a dedicated LLM processing rig for my house. But, this really needs to be the route in the future. Or just install a NPU on this speaker itself. Phones already do this, and it's been the standard since 2017.
If it keeps playback functionality only on LAN then it could be a good replacement for the Google Home units I have stashed around that constantly need to ping home even for basic local playback.
If you're talking about Google Home Mini or Google Nest Mini, there is a project on GitHub which replaces the PCB with an open alternative. The boards for the Home Mini are in production right now and Nest Mini is going to be the next.
Ooh, outstanding - I searched recently for whether it was possible to flash open firmware and that seemed to be a bust, but I never thought about a board replacement!
Thanks for the link, I'll be looking into this with great interest.
Edit: ah damn, just saw the price plus shipping to me, and it's a fair bit more than buying complete package other options. It's definitely an awesome initiative and this guy has done an admirable job, just might not be the best fit for me right now. I don't want to contribute to hardware waste, so will keep running what I've got instead of buying brand new speakers, but will keep an eye on this project for sure.
That's true, it's not cheap. But I have a bunch of them in the house which work great as speakers, but I have trouble integrating them properly in Home Assistant. And there are hardly any alternatives in the same form factor.
I had some trouble with them working with HA too, but a while back started running Lyrion Media Server as the backbone to all my local music streaming. Since then, they've all been pretty bulletproof (using the Lyrion/Squeezelite HA integration for all speakers, and no longer piecing together the integrations for the individual speaker types). YMMV, but FYI.
Does yours even work? I have a Polk Google nest capable sound bar and like 4 of the small nest mini speakers and the groups have always inconsistently worked and overall it works like utter shit.
Yeah I've got two sitting around and they work pretty well, but I'm annoyed that they need WAN access to stay working for playing my local media. Would prefer a WAN blackout at the router level to make sure that the microphone off toggle doesn't get "accidentally circumvented" by Google.
The Third Reality speaker also recently came out.
So now there are three. The Pine one, the Home Assistant one, and the third Reality one.
The home assistant one (Voice PE) works very well as a front for conversational home control, but poorly for music (although to be fair it does include a headphone jack).
This pine one seems larger, so maybe they’re doing better here with the speaker.
I don't need to ask HA what the weight of a cow is, or any other LLM stuff, I just want to control the home
So, I've been faffng around with a Pi Zero and voice assistant and it works fine - with HA doing all the processing in a VM elsewhere.
I agree these 3 devices look much nicer, so which do I go for?
Is this voice assistant something I can install on Android? I didn't realize there was any alternative to Google or Siri (or Bixby).
Mycroft (afaik now called "ovos"), and the home assistant internal one come to mind
I believe you can! You don't need this product though. See https://www.home-assistant.io/voice_control/android/
If you use Home Assistant you can configure the voice assistant to be the default one for android yes. Mine only does control and statuses for my local devices, can't ask questions or anything but that's probably configurable.
I just set two of these up. I'm not a HA guy so my HA server only has these assistants, my Sonos system, and Music Assistant set up.
I'm still fussing through setting up tasks, but it's slowly turning into Alexa Jarvis.
I have it set up to play, pause/stop music from music assistant with a really hacky yaml script.
It can tell me the weather forecast.
That's all local processing with Whispr(stt) and Piper (tts).
For those who want the "tell me the tallest mountain in the world" type of support, the answer is cloud llms. I tried ollama on my PC, but it's just too slow. I have an opencode go subscription and that works pretty well with deepseek-v4-flash. You can also use a standard open ai api key with one of the nano or micro models.
It works, not quite as good as Alexa, but it's not bad.
How does music sound on it?
top 50 comments