Was the year of the voice a mistake?

As we are already half an year in to the “Year of the voice” I am wondering if the effort has not been wasted and if this was the biggest topic to be handled by the team? I know this is not the only thing that is being worked on but this is one of the biggest effort consuming feature.
My main concern is how useful will be the voice feature without hot word detection. And it was made clear this will hard if not impossible to implement. If I have to press a button to activate the voice command why can’t I use an NFC tag or zigbee button to trigger a scene? Yes the functionality might be limited a bit compared to voice command but it will cover 90% of the use cases.
Just look and the What the heck section to see how many improvements and polish can still be done.

So basically share your opinion.

Who said that?
Rhasspy (from the same DEV providing “year of the voice”) had wake word working years ago on a RPI4.

https://rhasspy.readthedocs.io/en/latest/wake-word/

No it wasn’t.

We’re less than half way through the year and we have working local TTS and STT plus a whole bunch of intents in many, many languages.

What will the next 6 months bring?

1 Like

I think it wasn’t. Already in 2018, Google promised control in my native language. To date, this has not happened. I like this idea of HA and voice control as it is currently the only option to control my devices in my language.

As said already, Rhasspy can already solve that.
Rhasspy is a fully functionally local voice assistant already.
What is missing is really a good integration into HA, which is coming now, and a large enough userbase to entice hardware firms to come up with better solutions for satellite devices, microphones and the likes, which will happen with the HA userbase behind it.

Look at Rhasspy here:
https://rhasspy.readthedocs.io/en/latest/

As much as I agree with the rest of your post: I don’t think that the current size of the HA userbase will push especially hardware development companies to take a risk/chance there. As a side aspect definitely (“works with home assistant”), but as a spent decision to do specific R&D? Not sure about that…
But as long as stuff like e.g. the m5stack devices keeps being supported, or extends their product range with cool new stuff, I think we’ll be fine from the hardware side.

I second that it was a mistake, yes. But that’s just me…
My opinion:

  • HA is used by whole word, not only by english speaking people. No matter what you do english is and will be first and most supported of all languges. Slovenian language is very hard, and as such tough to implement. Currently slovenian IS on the list of suppported, but it doesn’t work - doesn’t recognize a single word.
  • i don’t talk with dead equipment - it looks “funny”, as it does when a person walks on the street and talks with itself (makes a phonecall with BT headphone…). It looks pretty darn fullish.
  • whole year will be spent for voice, ignoring all rest problems. It took more than one month to repair lack of scrooling in picture elements card, automations are still NOT better organized, breaking changes are made without any warnins before…etc… like someone said in another topic: from 2023.1 on all i do is repairing HA breaks after each update… even my folks at home begins to “laugh” at me, when i come home and every couple of days i get “HA is dead again…” this doesn’t work…that doesn’t work…
    A big problem i see here is also lack of communication or lack of listening to users from HA’s guys. It seems that no matter what people say, HA guys go and drive their own way…
3 Likes

good point, i also noted that is weird…

3 Likes

Incorrect.

Feel free to look at the list of bugfix labelled pull requests that have been merged: Pull requests · home-assistant/core · GitHub

Are they all voice related?

No they are not.

The HA userbase is actually quite large and if you make small electronic devices, then it is one to watch if you want success.
The move have already started.
If you look at all the ESP32S boards that suddenly comes with new features for audio, both input and output (ESP Audio DevKits I Espressif), then it is clear the chip makers are betting on that field for the next wave of developments.
Once the chips are there then the rest of the hardware will follow.
HA is not the only thing driving this move. The AI advances that have occurred have triggered a lot of new products with voice recognition, such as dictate devices for doctors, helpdesk robots, voice response systems for pone switch boards and so on.

The current version of the HA voice assistant is a bit handicapped, I agree, but I run Rhasspy and Rhasspys abilities will be the current aim for the HA voice assistant.
Rhasspy can do all the things needed for a proper voice assistant, so I am okay with a little handicapped version for now. It will come and I know Mike will not stop once the level of the current Rhasspy have been reached, but instead keep pushing the boundaries for what is possible with a local only voice assistant and the hardware development will help him surprise us all.

1 Like

Of course there are other things,too. It’s just… on the account of “voice” which has clear advantage over all, all other stuff is put behind and solved only after all voice related stuff is solved. Since time is always a factor( or lack of it) all other solutions come (too) late.
One thing i forgot to mention above: LD2410 in esphome - why a solution where we can change parameters from esphome is still not accepted, if it’s clear that it works? It’s still “waiting for appproval”.

In appr.3 years since i have HA i never spent that much time in correcting things as i do this year… sure, i spend a lot od time behind HA, but generally for improvements,new things etc…
Don’t get me wrong: i absolutely admire HA and guys behind it, it’s a great piece of software. But recent problems lead me to exploring (more stable?) alternatives, only to find out that there IS no real alternative to HA… so i guess my only hope remains that things will get better…

Incorrect again. A new developer was hired for this project. It takes little time away from the other devs and volunteers who work the day to day machinations.

Hmm…the opposite here…took me a lot of effort to configure everything the ‘right’ way, but once that was done it has been a smooth ride ever since :grin:

1 Like

Well, i guess that in this case you don’t have sql’s who went bye-bye when database changed without particular warning… (just one example…)
Changes are necesarry and good, and i understand and welcome them. It’s just… these changes could be better warned in advance, in a way like (useless and annyoing) esphome update warnings…

The voice activation topic was mentioned as hard in second year of the voice stream Year of the Voice - Chapter 2 - YouTube

I haven’t try the assistant you suggested but I doubt anything on Pi works supper great.

Hate SQLlite, switched to Mysql (on the host machine), it was running there anyway for other purposes…It now has approx 2 years of data, runs smooth…

Yeah, that was annoying, but very easy to turn off

You were warned about all the database updates in the release notes.

Read the release notes and make a full backup before updating.

Your other gripes (here and elsewhere) just boil down to HA operating the way the developers want it to rather than the way you want it to.

If you are unable to accept this perhaps you should try an alternative.

1 Like

Tom, the question is right, look around a bit:

https://www.kickstarter.com/projects/aiforeveryone/mycroft-mark-ii-the-open-voice-assistant/posts

I was not addressing a question. I was addressing incorrect statements. So I have no Idea what you are on about and no interest in clicking those irrelevant links.

Hence the title of the topic you are currently participating in…