I found it and read through it and your githubs and the other options. Awesome thank you! Installing now.
One thing I did notice in the write up on github is the lineage link in the writeup for linux reference the hyperlink to the recovery. I found the correct link at the bottom and it appears to be working fine. It’s line 6 in the linux instructions.
Will udpate when done but thank you!
No problem.
Someone else did the Linux instructions - I’ll have a look and correct.
Hi Im using wallpanel apk as a browser and using built-in camera detection to wake up display. After some thinkering with sensitivity i admit that the solution work almost flawlessly. The only problem is with very low light at night. I’ve remapped the mic mute button as a wake up screen. Inelegant but it works and the display is completely off when not used…
is there an example of how to setup automate?
Can you elaborate on the tinkering with sensitivity part?
He did a write up over at Setting up a 100% local smart speaker on an Android tablet using Tasker and Snowboy to handle wake word detection
A newer option is View-Assist with RtpMic, I haven’t tried that one yet.
I have and its solid.
Caveat though - its streaming the mic ALL the time.
It would be nice to have local wake word detection (I seem to remember local wake word detection could be done on a Pi Zero W2 - so hoping this is capable enough).
It’ll be too busy running android 
I’ve been running rtpmic with some of the View Assist screens (dashboards) for a while, helping test them and generally contributing where I can. It does work well and although rtpmic is constantly streaming, it doesn’t seem to negatively affect the device at all.
There is no local wakeword detection built into the home assistant app yet, you’ll have to run some other wakeword app that can handle it locally, the only one I’ve found for this so far is “hotword plugin”.
If you find another android app that can do it using tflite models please let me know though!
In Wallpanel, BTW I’m referring to this APP
Here’s a link to the doc : Getting Started | WallPanel
There’s a Camera Settings in the menu, once activated the motion detection I had to lower the Maximum Leniency to 20 ant the Minimum Luma to 200.
The thresholds I set are non-scientific, just tried some random values until I found something that works in my location on almost all light conditions.
I would have preferred that the TSV could be unsleeped at night with a double tap on the LCD but I’ll have to cope, I’m not using the TSV to turn on the lights so it’s not a serious problem for me.
You can enable the volume buttons to wake the screen in system settings. A little easier to press one of those buttons than use the mic mute or stiff camera privacy slider.
Hi,
Yes i tough about that but I’ve chosen the mic switch because I wanted to leave the volume buttons as they are in case in the future I’ll need to play some media on the TSV.
In any case thanks for the suggestion!
I used this tutorial on a Pi zero W and it was solid.
I’m hugely dunning-Kruger-ing myself here, but if it can run on a pi zero w then surely this Lenovo has the chops? @synesthesiam (sorry to cold summon you, think of it more like a “hey buddy, any thought on this?”). Please don’t let your creativeness be impeded by my poorly thought out technical waffling - but….
Automate seems like a good option, I wonder if the delay would be too slow?
Something like
Automate hears wake word, which enables the micRTP stream. Then it would need some kind of trigger to disable the stream again after.
The micRTP app could always be running, it would just need pausing somehow so could rapidly resume state on a wake word?
Any other android apps that could use the mic for wake word detection and trigger the stream to HA to use the awesome Wyoming pipeline (is flattery working 
)?
It makes sense to use Stream Assist for the audio pipeline server side.
So far LOVE these devices - I have 5 now and a bunch of friends bought some too.
Edit: picovoice
Regarding audio streaming for these devices - does any of the panel apps? Fully kiosk or wall panel, have any interest in using its mic capability and adding a wake word plugin?
Would be kind cool if the kiosk/panel app provides a streaming source as well as a speaker source to start. And then added wake word detection to kick off the audio streaming and pause it when the session finishes?
Edit: there’s a guide here:
I’ll have a look and see if I could fork the code, and try with this or RTPmic - but I fear I am WAYYYYY out of my depth
Is there a way to do that on the Android 11 ROM?
It’s a system option so maybe but I don’t have A11 any more.
Has anyone experimented with a native screensaver instead of the ones built into wallpanel/FK? Im trying to decide if this would work: Releases · 3rob3/ImmichFrame · GitHub vs my half s$$ed implemenation of it as a HA dashboard via the google photos integration. I use both gphotos and immich but gphotos api has a request restriction so if i have multiple devices showing a slideshow they end up erroring out and showing a missing image icon instead of the photos. There is also this integration GitHub - outadoc/immich-home-assistant: Home Assistant component to display random pictures from your Immich library. but it doesnt have a configurable interval (just fixed to 5 minutes) and it also doesnt have a few nice things that the gphotos integration has like aspect ratio cropping settings and stuff. Was thinking about just trying that immichframe as an apk for the screensaver and then flip to wallpanel/fk when someone touches the screen or something.
The hardware itself is more than capable of running local wakeword detection software. And the HA companion app provides an easy way to connect Automate to the assist pipeline (it can be called directly as a service).
There’s technically nothing preventing you running wyoming satellite on these devices, android is Linux based after all. But even with root access I lack the technical skills to get it working.
My devices running hotword plugin have a very short delay, the device recognizes the wakeword and is ready to take a command with a 1-2 seconds. Overall I’ve been very happy with it’s performance, my only issue is that you can’t use the same wakewords as other wyoming satellites, because hotword plugin doesn’t use the same software for its detection. You can see it working in the video I posted here: View Assist - Visual feedback for Assist voice assistant on an Android tablet (Install info provided on Wiki) - #20 by Endlessvoid
2 quick notes on picovoice, cause I looked into this as an option as well: