this post was submitted on 21 Jan 2026
24 points (100.0% liked)
homeassistant
17594 readers
26 users here now
Home Assistant is open source home automation that puts local control and privacy first.
Powered by a worldwide community of tinkerers and DIY enthusiasts.
Home Assistant can be self-installed on ProxMox, Raspberry Pi, or even purchased pre-installed: Home Assistant: Installation
Discussion of Home-Assistant adjacent topics is absolutely fine, within reason.
If you're not sure, DM @GreatAlbatross@feddit.uk
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I'm currently doing it the other way around. Assemble multiple satellites and spread them through the house. With upcoming Sendspin and Music Assistant this might do whole house audio soon. But I don't own a Voice PE. I just bought some microcontrollers plus MAX98357A codec/amplifiers and connect them to random old speakers I have in my e-waste / upcycling bin. The one thing with an 3.5" audio jack might just go into the preexisting soundbar or stereo in the livingroom.
Are you doing anything special for the microphone(s) and noise cancellation? I'm leaning towards making my own satellites as I'd like a half a dozen or so and the PEs are more than I want to spend on the project.
Uh, noise cancellation is hard. First of all, the audio pipeline currently isn't able to resample the microphones, so mic and output need to be connected to separate i2s buses, or it won't work simultaneously in the first place.
And then I had some luck with the microwakeword component. It often triggers correctly even with noise in the background. And I have an automation that mutes all media players and the TV when the wake word is triggered. That's my "noise cancelling".
I think more elaborate noise cancelling is going to require some dedicated hardware (or maybe some proprietary ESP-ADF functions) and a microphone array. But that's probably as expensive as an Voice PE?!
I'm not in a good place with the voice assistant anyway. Don't own a graphics card. So it's slow. And Whisper never gets all the words right for me. So it's down to the speech-to-phrase addon. And that seems to be broken as of now. At least I get more connection errors than commands through. I think I'm going to do the Sendspin media player first. And then maybe add a microphone and voice assistant later.
I do OK with faster-whisper for transcribing, and I built a wyoming container for pocket-tts that does pretty good local TTS but I am running it as a docker on a ryzen machine (no gpu) so YMMV. Pocket-TTS seems better than Piper IMO, it's certainly faster for local TTS if all you're using is CPU.
I've been looking for a MCU that has enough oomph to do some noise cancellation onboard and I ordered up a couple of these in the hopes that the onboard NPU would be useful for that. It also has a speaker output and onboard mic. Price was right for 8GB of EMMC and 256MB of ram.
Nice, thanks for the link! I wasn't aware of that. Sadly as with all shiny new things it doesn't fit all my requirements... I'd really like to speak to my house in my native language. But I figure English will do. I'm gonna try that.
Not sure if an ESP32-S3 is fast enough for more advanced DSP plus the rest of an voice assistant. At least I found some ESP32 libraries with noise reduction, echo cancellation... There is the ESP-ADF and a project called ESP32-SpeexDSP. But I didn't try that yet. The Rockckip / Luckfox development board looks nice as well. A Cortex-A7 and a few hundred megabytes of memory might come in handy. And whatever the NPU does. But I don't have a clue what kind of software and libraries we got for embedded Linux or custom processing units.
Anyway. I think the production-grade stuff mostly uses multiple microphones and a combination of beamforming and echo cancellation. I got 4 inmp441 microphones here. But I lack the software/libraries to tinker with that kind of signal processing.