alarms and timers

[cinder stream]

I'm thinking about making the switch from using Alexa to the home assistant voice assistant, and the main thing I use Alexa for is setting alarms and timers. Is it possible to say "hey Jarvis, set an alarm for 7 am" ?

[worn ember]

I do it. It requires some preparation, but works flawlessly.

However, think twice about satellites.

[pearl canopy]

Why “think twice about satellites”? Because there aren’t any good retail versions? Hard to maintain? They don’t work reliably?

[worn ember]

There's no good ready-made product as of now. Several reportedly in development, the PE from Nabu being the best. But no details.
Solutions, that are available right now (I tried):

• Wyoming-satellite, liux-baded device. I tried Pi Zero 2W +; Respeaker 2-mic hat, but it didn't work good for me, too bad with speech recognition...
• esp32-s3-box based. Decent microphones, but sometimes still doesn't hear me from same room. Awful speaker.
• other ESP32 devices with single i2s microphone and DAC. Can be enough for small quiet room, but don't work in noisy or big rooms at all.
• (one I find best right now) ESP32S3 + Seeed Respeaker Lite. Very experimental, but has 2-mic array, ruled by XMOS chip, which works pretty well. Still not ideal, but at least satisfies my family (which I couldn't say a out others).

All of them require tinkering and most require soldering. 🙂 and none can replace Alexa in convenience so far.

[cinder stream]

I'll look into satellites, I'm new to the whole scene and I'm probably jumping to far into the deep end with my project right now

[worn ember]

Well, it's very interesting rabbit hole, if you're down to tinkering and don't expect same level of experience as you have with proprietary systems, evolving for decade. I can say that my setup satisfies me pretty much. I'm working on decent enclosure for satellites now, and after that will be ready to say "I'm done with voice hardware in my house". 🙂

[pearl canopy]

Yeah I’m building a new home and pulling my own low voltage wiring so was hoping to find a great Alexa alternative but suppose I could use them in my ceiling with HTD’s Alexa kit if needed.

Alternatively, I wish I could find a whole home microphone solution (e.g. dozens of in-ceiling microphones dispersed throughout my house connected to a mixer -> connected to a satellite or HA itself listening for wake word.

I’ll already have whole home speakers spread throughout my home via HTD’s Lync Amps

[pearl canopy]

About how far away can you be with the Seeed speaker lite and the mic still pick up your wake word? Does it work decent say 5-10’ away if music is playing in the same area?

[worn ember]

Whole-home microphone isn't the best solution though, because you won't be able to distinguish the room. Voice assistant satellites, on the other hand, have area context, so you can say "turn on top light" or "play music" without explicitly saying the room. It doesn't seem very convenient at first, but it really is.

[worn ember]

Yes it does. I actually can successfully use it from 4-5 meters with working TV. I just need to raise my voice a bit more I had to with Alexa.

[pearl canopy]

Where is that satellite physically? Do you think it would work if installed in a 12’ ceiling?

[worn ember]

Probably. Need to test. But Seeed devs said they will be thinking on adjustable gain for microphones - that would be very helpful for long-range work.

[pearl canopy]

Do you happen to have the exact parts you purchased and a tutorial/video on how to setup this satellite you believe works the best?

[worn ember]

The parts are here: https://wiki.seeedstudio.com/xiao_respeaker/ i ordered kit.
No tutorial, but here's the thread on it: https://community.home-assistant.io/t/respeaker-lite-new-seeed-studio-voice-assistant-development-kit-hardware-combine-esp32-with-xmos-xu316-dsp-chip-for-advanced-audio-processing-as-a-esphome-based-home-assistant-assist-satellite-voice-devkit/756944

[copper drift]

Sorry for the ping, just looking for more info about the weeds of delving into alarms set via HA's assistant. I've been trying to get it to change the time of a helper, in which I can then set additional automations up from there, but im not getting anywhere with this on the base assistant, nor the gpt aided assistant. Presuming there's some translation that needs to go into saying "10am" to the format it needs to change the helper etc?

I also have timer helpers that I'm looking to be able to ask to set for x amount of time, then starting the timer with it being visible on the cards that I have on the dashboard (which I also can't get working for some reason)

[worn ember]

Hi! Yeah I have it working.. IDK if it will work on your language, I have it in English.
Sorry, I'm AFK and can't copy code from phone. Will try to not forget pasting it after getting home. If you can ping me in 12 hours, would be great.

[elder geyser]

I have been testing various amplification factors for my microphones using your project. The amplification is just a shift left of the microphone data. The volume of the recording does change as verified by reviewing the debug recordings. Part of my lower microphone volume could be due to my enclosure design.

      id: nabu_mic_mww
      amplify_shift: 1
    channel_1:
      id: nabu_mic_va
      amplify_shift: 3

[worn ember]

Nice! So you confirmed amplification working, and fine-tuned for your enclosure?

[elder geyser]

I confirmed the amplification of the request recording by adding the
debug_recording_dir: /share/assist_pipeline to the configuration.yaml file.
The wakeword is not recorded in the recording files. I am still experimenting. My enclosure is a vertical cylinder with the Respeaker-lite board mounted about 0.5 inches from the top. Your design is more directed at the person speaking.

I am just surmising that part of my issue is the enclosure design. I know very little about microphone sensitivity/pickup and how it is influenced by the case design. I am also using "Hey Jarvis" for the wakeword. I have also been experimenting with the wakeword probability set in the 0.96 to 0.99 range to limit the false wake when watching TV. The false wake events bug my wife the most because she watches more TV than I do.

[worn ember]

Yeah Jarvis wake word is way worse than Nabu. Do you have VAD model as well? Unlikely that it will help much, but without it situation is even worse.

[elder geyser]

I personally did not have much luck with "Ok Nabu" . Maybe it is my midwestern accent and I should revisit using Nabu wakeword. I also use the Jarvis response "At your service" for the wake_word_triggered_sound sound and have a Jarvis voice for my Piper TTS response.

[worn ember]

Probably because of enclosure?