Voice Preview Edition - Pain points survey

Configuration Voice Assistant
1

Hey hello! Its Laura from the Open Home Foundation. I’m trying to gather all the most relevant pain points everyone has when talking about Voice, more specifically using the Nabucasa hardware VPE, and I need to do so in context of the setup (type of hardware running HA, if you are using AI or not, etc).

To avoid getting it too chaotic in comment threads I created a google form for this. If you have been using VPE (even slightly) and want to contribute with your pain points, it will be super helpful. Here it is: https://forms.gle/9qubAWUagcfnRToz6

I am aware a lot of the things shared are going to be things you already shared, but having all the feedback in context with the setup gives us an organized way to access all this information all at once.

Thanks again!!
Lau.

3 Likes
2

That was it ? Short answers only, on a single web page form which probably times out after 20 minutes if a user tries to give a longer considered response.

My situation is a VPE in living room, and a LVA (running on RasPi 3B with reSpeaker 2-mic HAT) in my study. I was previously using Rhasspy for several years. I tried to do a minimal setup due to focus on another project, so using the HA Cloud.

I swapped the primary language on VPE to Italian for my partner, with Okay Nabu as the wake word for the IT assistant, and I have used “Hey Mycroft” to wake the English assistant.

  • I setup a “good morning” automation to open the blind, and was easily able to add the Italian trigger word :+1: I had the automation responding with “Have a nice day” … but it is not possible within the automation to detect which language is being used - thus two separate automations are required. Fortunately my Good Morning is rather simple … but having to maintain two copies of multiple automations could become a nightmare.
  • Two wakewords means twice as many false positives.
  • I have found that it is unusual for the VPE to respond to Mycroft, instead the LVA in another room responds and (because of the distance and simple mics) often does not recognise the command. About half the time both devices simply ignore my command.

Consequently neither of us use voice commands much any more :frowning_face:

To be honest I haven’t put in much effort to troubleshoot VPE due to (a) wanting to finish off my greenhouse project before getting distracted, and (b) low resilience due to old age

But for me the big issue is users’ perception.

I personally have worked as a programmer but customer facing, and learnt to translate between user terminology and understanding, and technical issues and staff. I appreciate the developers point-of-view, but also that of users who get the impression that their concerns/issues are being ignored.

I see the basic nature of Open Source projects (including HA/ESPHome/Voice Assist) as “by developers for developers” with the assumption that everyone else is desperate to learn the skills to be a developer. Discord and Github are exclusively the province of developers - even attempting to report an issue requires a fair amount of expertise - and responses are often cryptic references to things no user would understand. Assuming the issue isn’t just ignored until it is “completed” as stale.

OTOH, “users” report issues and raise queries in the HA Community … where they are answered with other users’ opinions rather than knowledge.

Rather than try to force users to become experienced developers; or for developers to waste time responding to new users repeatedly asking the same questions … I suggest a two level system where
(a) users report possible issues to a helpdesk type application,
(b) Nabu Casa employs staff to moderate the helpdesk, triage issues, answer simple/repeated questions, track issues, and vet genuine issues by asking for additional information, etc before
(c) creating a github issue for developers.

Thank you for listening.

1 Like