Home Assistant Voice

The Sonos API is already open.  Any company wishing to integrate with Sonos can do so today:

https://developer.sonos.com/s/?language=en_US

Does the API also enable access to the microphone?

No. But that isn't what was specifically asked.  “Home Assistant Voice” is HA’s own  voice assistant, and thus it’s an interface for HA.  HA is using the API to control Sonos.

That said, it does seem that the topic is prone to miscommunication as to what exaclty people are asking for with this feature request.

There is a misunderstanding here.

The only thing that is required is to open up the smart speaker microphone as an entity in Home Assistant. No other voice features, AI, Assist or anything else is required. Just have the microphone visible and usable as an entity in Home Assistant, that’s all.

I can control my Home Assistant through my Android phone microphone, using only local resources (Piper, Whisperr, openwakeword, Ollama), using Home Assistant’s Assist. I would like to use the Sonos microphone to do the same. Just take my voice and push it to Home Assistant through that microphone, that’s all.

Well, the Sonos would need to then send the data to the Home Assistant servers, and then parse any relevant data that comes back, too. It isn’t as easy as ‘just opening the mic’

I never said that an integration that enables SVC to control HA would require AI.  I stated that Sonos maybe should focus on AI over integrations like this.

edit: I should have stated “an integration that allows HA Assist to use the Sonos speakers mics.  No SVC involved.

It doesn’t need to parse any data that comes back.

It needs to send the voice to a STT service, of course, for example using Wyoming Protocol, but that is literally all. Home Assistant can handle everything else.

I've been considering ways of moving away from Alexa / Google and one of the things that holds me back is a lack of support for features that I've come to rely on - namely home control by voice.

Where there is a Sonos speaker with a mic, I don't want to have to wire her another device with a microphone to be able to maintain current functionaliry.

I stumbled across this thread only to be frankly amazed at how many people are putting forward arguments for Sonos NOT supporting tighter integration with Home Assistant…

I mean… Why?! 

Sonos have their big person pants on, can do their own research and can also consider that certain services without huge market presence ‘coukd’ have potential, especially with their help, and especially as a company that has been through some trust issues on their own part …

The statistics quoted re use of e.g. Alexa and Google are pointless. They are that high because they are easily accessible. People don't really choose voice services in the conventional sense.

Were SONOS to contact and work on integration with Home Assistant, to be able to use a Sonos speaker to communicate with and control a home assistant setup that would be offering folk and alternative option. It seems like an exciting prospect.

I fully appreciate my own use case is incredibly niche - I have a voice disability, my partner doesn't. They still use their voice, I have worked out ways of interacting with my systems that don't require my voice now  - buttons, scripts and automations. Home Assistant has been brilliant for that and works already. But if Sonos could actually as a way to provide voice inputs for Home Assistant that would be even better, far more readily allow for more stable scripts and automations whilst preserving my partner’s ability to do all that with their voice still. It would mean not having to rely on so many third parties.

Alexa and Google don't really give a damn about accessibility. Home Assistant allows for far more flexibility beyond what any individual in a high castle would consider might ever be ‘useful’ to someone.

It just seems weird frankly that anyone would discourage that. 

After 7 months, lots of reply. Still no answer from Sonos, neither on supporting an open standard that could be a good move on the eyes of the customers, also on the future of google home assistant, which is getting worse and worse waiting for the LLM version.

How many developers do you have Sonos? This looks like a feature that you can implement in a few iterations.

Inside a Sonos One you typically have:

CPU

• Cortex-A7 / ARMv7 (Gen 1)

• Upgraded ARM SoC (Gen 2, faster CPU & more RAM)

Memory

• 256–512 MB RAM depending on model

• 4–8 GB flash

DSP

• Dedicated DSP for:

  • echo cancellation

  • noise suppression

  • far-field mic processing

  • beamforming

This is more than enough to run:

• wake-word models

• streaming ASR pipelines

• local inference at the scale of Wyoming Home Assistant voice modules

• or even a small offline speech model

A Raspberry Pi Zero can handle this.
A Sonos One is far more powerful, especially with its DSP chain.

Ended up here while doing some research for my next home audio upgrade, Sonos ticks almost all the boxes, except for this particular requirement.

As for all the comments about lack of market, that's kind of irrelevant. Development effort is relatively minor, and in an atmosphere where Sonos is desperate to regain trust with their customer base, implementing such a privacy-centric feature would generate a TON of positive buzz. They'd have their ROI in marketing alone.