Yes and no. That was the beginning of my question these days.
As the forum rules ask to not ping, @, or DM users, how should I discuss is with the language leaders? And Tom said, I should ask here, where (my gut feeling) the language leaders are unfortunately most probably not participating.
At the end of the day I like all the efforts going into the Year of the Voice. But unfortunately, everything is build on each other, so hardware voice recognition is using STT and this text is then running into the text engine. But if this ground and foundation is not working as expected or described, how should all the other uplifting iterations work?
So my gut feeling is, that the text engine have to work as best as it can and feedback, what is not working should be somewhere centrally discussed. Ofc I would learn syntaxes and a lot of things could lead to not working sentences, like wrong devices, wrong device classes, wrong attributes, wrong setting wrong rules wrong calling and running the rules. But how should all this be improved without discussion or feedback possibilities.
Then I gave the rule (and still I’m sure that the language is irrelevant, these are only filling words and patterns) and how I call the rule and asked if it shouldn’t work or where perhaps is the problem, but this was not enough. Don’t understand but fair enough.
For me, ask for # of status of windows open is not working, ask for temperatures is not working, etc. etc. Will I go on to use it and? I don’t know yet.
Sounds like what I am working through. Make sure all your areas are properly set. If you have punctuation or other German accented letters then try an alias without and see if that works
The blog post indicates “that is optimized for running on a Raspberry Pi 4” - however when I go to the add on store on my raspberry pi 4 installation, I see “This add-on is not compatible with the processor of your device or the operating system you have installed on your device.” - just wonder how I can get this up and running?
This is great progress.
Talking of voice :), has anyone created a voip for HA on Pi such that I can call in from abroad and have it “break out” to the home country’s landline or an attached phone or sim device?
The ideal scenario would be an integration on HA and on the app, when abroad can call a conventional landline through HA and home landline or mobile using a touchpad remotely.
The reason for this is that when dealing with companies in your home country, they only advertise landline contact numbers. They don’t offer a voip number or WhatsApp yet . Although they often use your mobile number for faster ID, they have non id by mobile options. Calling home from abroad to do admin costs a lot using voip providers.
Thanks for any ideas.
Is there any perspective on being able to accelerate whisper (and piper) using a Google Coral accelerator? Seems like an ideal setup for a HA Yellow, which is more or less unusable right now, at least for non-en languages.
For the IP and port (in the Wyoming integration), I am using the Pi’s IP and piper's port (10200). Is this related to the fact that I am not running HA over https locally? If this is the case, how do I convert to https?
should not have had a space between the -- and voice=
I fixed that, restarted and now piper's docker show the following:
INFO:__main__:Downloading en-gb-southern_english_female-low to /data
INFO:__main__:Ready
ERROR:asyncio:Task exception was never retrieved
future: <Task finished name='Task-6' coro=<AsyncEventHandler.run() done, defined at /usr/local/lib/python3.9/dist-packages/wyoming/server.py:26> exception=JSONDecodeError('Expecting value: line 1 column 1 (char 0)')>
Traceback (most recent call last):
File "/usr/local/lib/python3.9/dist-packages/wyoming/server.py", line 28, in run
event = await async_read_event(self.reader)
File "/usr/local/lib/python3.9/dist-packages/wyoming/event.py", line 48, in async_read_event
event_dict = json.loads(json_line)
File "/usr/lib/python3.9/json/__init__.py", line 346, in loads
return _default_decoder.decode(s)
File "/usr/lib/python3.9/json/decoder.py", line 337, in decode
obj, end = self.raw_decode(s, idx=_w(s, 0).end())
File "/usr/lib/python3.9/json/decoder.py", line 355, in raw_decode
raise JSONDecodeError("Expecting value", s, err.value) from None
json.decoder.JSONDecodeError: Expecting value: line 1 column 1 (char 0)
The entities exposed to the assistants (in my case Google assistant) that we set up via exposed domains in the cloud tab - they are all gone now? There are only a few entities still exposed - was that supposed to happen? There are not even filters/domains to select anymore.
Hello. This has been a great blog post and a great release all together. I have installed/configured Whisper nd Piper as HA add-ons with all the standard settings for english language useng the whisper tiny-int8 model. My HA intallation is on a VM on proxmox running on an i7 with all 8 cores accessible. My main question revolves around performance. According to the livestream as well as info found online, the stt engine should run in sub-second response times for small commands however when i look into Debug of the assist pipeline I see it taking 4-5 seconds minimum. Are there any more info regarding performace or maybe some guidelines?
Yeah I am seeing similar when I have put whisper on my synology nas to try and keep the load off my RPi. I’m using tiny-int8 as recommended. When I did the first one or two commands using HA cloud it was super fast but now takes 5-7 secs to do the stt. So I have tried both but they are both about the same
I saw this error when I tried to install Wyoming first without Piper and Whisper installed first.
Are both of those installed?
Second, I was seeing issues when I tried to access the system via the direct IP route on our mobile devices. PROBLEM: 192.168.1.1: 8123 WORKS FINE: https://mydomain.homeassistant.org
That problem manifested itself as the assist button does NOT show up at all.
