My thoughts are, as stated in the YouTube stream, NabuCasa is very unlikely to release different hardware versions in the near future.
So people might stop holding their breath for this.
On the other hand, external parties can, and have already begun to, develop more feature intensive products based on the existing PE platform, since all specifications are open source.
So maybe there will be a version integrated into a microwave oven with 250W speakers and wheels soon.
You could see it this way:
The PVE is a WiFi microphone for assist and it comes with a low end speaker as a bonus… Taking the speaker out would not save much in space and cost.
Personally I only listen to music on my main stereo system at home so the ‘speech’ quality speaker of the PVE is exactly what I need. If you need more, just plug in an external speaker to it. Volume can still be controlled in HA.
Modularity is key for such a general use device. Otherwise, there are many DIY ways to match one’s personal needs.
Can’t see why it would not be possible in an automation. Just use the “assist in progress” sensor as a trigger. Lower the volume of the Sonos for x sec and set it back to what it was before (need an extra variable/input_number for this)
The jack is an output, not an input
And the vpe uses the media player component so you have control over the volume.
Also, i hear so many things about the internal speaker but remember this is presented as a voice client not as a netwerk speaker. It can be but then you have to plug in you own speaker.
For hearing speech clearly you do not need a full range speaker. Even the speaker of a phone sounds crap.
I’d like to see microphone-only devices too. But it doesn’t seem likely, having the decision to have HA send the audio response to the voice satellite, and the satellite redirecting it elsewhere (if configured in yaml). To me, much better solution would be to have voice satellites as HA devices, which have 2 configurable fields - microphone and speaker. So you could create new voice satellite in HA, specifying which mic and which speaker this device will use. Then HA could easily lower the speaker volume if the microphone detected wake word. It would also save the sound redirecting via the satellite, as HA would send it directly to the configured speaker. Current devices, such as this new Home Assistant Voice would create the device in HA, configured with it’s own mic and speaker, but allowing user to change the config - which means it would allow users to easily set up different speaker in GUI.
That is all correct, it is FINE if you ONLY want voice commands.
I don’t have my own speakers OTHER than complete google minis currently I would like to replace… About 69% of 252 voters of this forum responded they expect their voice assistant to play music.
So we would need to purchase speakers… or it would be nice if there was a variant of this device with a equivalent speaker to a google mini / alexa dot / home pod mini.
Here’s a tip. A reasonable to good speaker can be bought for $10-15,- however just have a look at a garage sale or a 2nd hand shop. I got one for just $2,-
Now this remark to will get me roasted but i will make it anyway.
69% of the votes want to play music through the VPE. It just can do that. Nowhere is stated that this must be HiFi… well it is HiFi and steteo through the external connection. Not via the internal speaker. That one is just mono.
Completely agree with you. Not everyone has the same definition of hifi anyway. A good speaker for someone might not be good enough for others.
For instance I will never listen to music on an alexa/nest or whatever little speaker box, I personally find it inaudible. So I fully embrace the flexibility of having a competent enough DAC in the VPE and plugging in whatever I consider a good hifi amp/speaker(s).
My preference would be to spend about £20-£40 on ebay or elsewhere on a pair of ‘bookshelf’ speakers.
I have a pair of Kenwood ES from the late 80’s when they where still made in Japan and are in grade A condition.
A Samsung subwoofer off a surround system with its own 100watt mono and 160watt stereo amp module boards on hex pillars epoxied to the back of the subwoofer with a PiZero2 running Snapcast as the gpio mutes when not playing audio and both boards have standby/mute connections.
The Kenwood are 75watt rated and the amps like all amps are rated at a horrible THD (Total Harmonic Distortion) of 10% at some stupid ohms like 2 ohms.
I just set the gain pots so they are running about 30/40 watts prob and far away from the terrible quality slope off many amps have nearing max specification.
The snapcast server runs on my workstation as a single guy in a flat, whole house multiroom synced audio is not needed and could of had a Phono connection but geekery got me interested on howto do a opensource Sonos like system that is snapcast.
Likely a AI/Voice server would also be a central snapcast server and would just pipe my computer to one of its stream inputs so any client can connect if wished.
Linux has a plethora of highly optimised C audio libs from ALSA , Pipewire, Gstreamer and multiroom audio servers such as Snapcast or Squeezelite via LMS server.
Why the wheel is being reinvented in Python as P2P is bemusing, but hey there is a wide range of what people see as requirements.
If great production tested opensource has already existed for many years, with many addons and hardware implementations, why not share and support those great opensource userbases.
It seems like meglomania where HA is ignoring such great software just to provide own brands, with claims they are open standards as they are purely what HA devs did without consultation…
From HiFi-Amped is a Raspberry Pi Audio HAT with dual DACs and class-D amplifiers - CNX Software Pi Hats for £25 to even cheaper ESP32 Squeezelite boards they do GitHub - sonocotta/esp32-audio-dock: Audio docks for ESP32 mini (ESP32, ESP32C3, ESP32S2 and ESP8266 mini modules from Wemos) and others like https://raspiaudio.com/
There are already 2 well establised opensoure ecosystems for multiroom synced wireless audio systems with numerous vendors and addons from various systems.
That are currently far feature rich than the new inhouse HA system …
Squeezelite/LMS a lighter system suitable for ESP32 or Snapcast which is heavier but a Sonos competitor for a PiZero and upwards.
GitHub - LMS-Community/slimserver: Server for Squeezebox and compatible players. This server is also called Lyrion Music Server. to GitHub - sle118/squeezelite-esp32: ESP32 Music streaming based on Squeezelite, with support for multi-room sync, AirPlay, Bluetooth, Hardware buttons, display and more to GitHub - badaix/snapcast: Synchronous multiroom audio player
Why great opensource is ignoring great opensource to create ownbrands to me isn’t really an opensource process but hey, each to thier own.
I have been using these modules as they do have standby’s and getting them to standby and not cause hum and save energy is a pretty esential feature nowadays.
The TDA7498E Audio Power Amplifier Board Btl Module for Computer Vehicle Car | eBay TDA7498E boards like this are plentifull with both mute and standby.
DC 12V-24V TPA3116 Mono Channel Digital Power Audio Amplifier Board BTL Out 100W | eBay bridged mono prob make good woofer boards with a mute.
Near all the boards from Sure Electronics (Wondom) have standby and mute WONDOM OFFICIAL SHOP - Amplifier Board - Sure Electronics - ADAU1701 - 18650 charger - Sigmastudio and they are a step up in quality often with more filtering, heavier weight PCBs…
With a google you can find many but standby and mute IO so theycan be controlled is a big consideration, but they are often equal in quality to much consumer audio and often very cheap especially on aliexpress.
Can anyone recommend a ceiling mount (good quality) speaker compliant with this 3.5mm jack ?
Partner isn’t going to approve going from a tiny google mini in each room to a new mini size devices PLUS a set of old PC speakers… The space commitment for that solution is about 3x at least.
It is FINE for some… I am just saying I hope a Likel for like or at least similar size solution is made. I find the VOICE ONLY solution a little odd that is all.
I agree and as mentioned in the reSpeaker Lite thread, you and others likeminded (and that include myself too) will probably instead be interested in following this Onju Voice feature request discussion about making updated custom PCB (Printed Citcuit Board) to be semi-DIY drop-in replacement for the original PCB inside the Google Home Mini and Google Nest Mini, so check this out:
It shoud in theoey also not be as expensive to retrofit your existing Google Home Mini and Google Nest Mini speakers with just a new custom PCBs compared to buying brand new voice assistant hardware.
As least that would ve a nice short term solution, as in the long term I will probably buy new speakers sooner or later, but it would cost a lot more to replace them all in one go.
Ya, the dev doesn’t have access to the first gen units so fingers crossed, but I might not ever see that retrofit for the original google mini.
I would be willing to pay more per device for a better speaker even if the device was a little bigger like a full size alexa, just to have the one device, one cable simplicity per room.
Google / Apple / Amazon all have tiny “low end speakers” and bigger better speaker units… Would be nice to see the same here… The difference however is that the current presented LOW is even lower in audio quality.
Oddly enough the presented solution is probably best for those with HIGH audio expectations due to the external jack.
Remember that the currently available product is labeled as a Preview Edition, which means they are likely reading the thread and taking our feedback into account for the final product. This means that the final product may just have a better speaker than the current one does
I hope they are… or with the open design I hope someone is…
I was thinking about trying to build something that the PE can sit in top of and uses the grove port for an I2C DAC in a speaker enclosure, as I too believe for maximum benefit these devices should have good quality speakers for music, even just to hear the voice responses, these tiny speakers on the PE or S3 Box 3 are puny, and though they might be ok for some, in my home they’re inpossible to hear over the kids playing, or the TV, or from another room, or if music is playing from another device without ducking.
For now, I swapped out the Pi in my home office running Moode with the Voice PE; it was connected to a set of studio reference speakers via 3.5mm so was a quick swap.
The audio quality from the PE is much better than I was expecting on the 3.5mm out, it’s excellent frankly.
As an added bonus, even with rock/metal playing pretty loudly, it can hear me mutter “ok, nabu” under my voice from 6 feet away, though it doesn’t pick up any of the other wake words I’ve configured, so they don’t seem to be as well tuned as ok, nabu.
Now I’m just trying to work on getting ollama to do anything resembling controlling my home properly without tuning off all of the lights because I asked it to play music or open the blinds, or as it did passive-aggresively last night while I was watching a show with my wife, turned off all of the lights in the living room to create a better ambiance because it false-activated then heard what was being said on the TV. 
I really love the concept of controlling amplifer/speaker power via the Grove-port for two reasons:
With a proper setup it could allow the solution to fully power down the amplifier so it does not use any or at least much less power when not playing music.
Using a high-quality Class-D amplifier (digital electronic switching amplifier) not only uses less power but can be completly silent and without hum when powered on and not playing anything, so have a very good Home Approval Factor.
As long as you are not an audiophibe then think those two factors are extremly important, since in comparison my existing Hi-Fi amplifiers draw about 60-100Watt each even if nothing is playing, and also my powered speakers draw a lot too even idle.
Currently I use those with Chromecast Audio dongles and get around the constant power draw by using a smart power-plug to the amplifiers/speakers to only power them on via automation (which is a pain).
Somewhat off-topic:
Your amp shouldn’t hum. 
Class AB amps pull power according to the volume that it’s set to (the amount of amplification), regardless of an input signal, so just so you know: turning down the volume on a power amp will save you power.
You’re right about class D amps. They are overall more energy efficient.
100W with nothing playing! Something is wrong with your amp
The modern D class amp silicon are really good and with a minimal amount of components produce really good efficiencies 95%.
I think mute on many you just need to take the input to ground and very little power is pulled.
I guess you don’t need anything but an event when no sound has been recieved for x-time and HA can turnoff a DP relay to your speakers with a short delay and then another relay of the power to the amp.
A classD amp with a mute and standby would likely be better as the inrush of switching an amp like that often is a harsh test of any psu and amp.
They are really cheap https://vi.aliexpress.com/item/1005008060712959.html and often cost less than the power brick you need for them.
Maybe even tidier if there was a way to combine the Grove as a 6 pin with the L/R of the 3.5mm so its just a single cable.