Yeah sorry, I was thinking of online assistants. My mistake…
Did you try running picoTTS in its own docker container? Is picoTTS the only library you’ve tried? I’m the one in a bubble?
I didn’t try to run picoTTS in a docker because it’s not made for it (you can access it only locally in your system).
I also look specially for a solution which runs (locally) on/in hassio and not some of the 20-x solutions which runs on other peoples computer (cloud) or which need extra hardware or a dedicated host.
While I found out about marryTTS which is supported by home assistant it looks like it’s not maintained and lacks a add-on for hassio.
Still 2019, still no local tts for hassio available (AFAIK)
1 Like
There is a issue out on github that picotts is indeed completely broken with hassio (up to the most recent version 0.100…
No local, offline & private tts for hassio right now 
I’ve recently managed to compile nanotts from here:
Under Raspbian I had install a few extras
sudo apt-get install build-essential libtool
but it does in fact build OK. If you need local offline TTS I’d recommend Raspbian or Debian, but if you are a Docker wizard you may be able build nanoTTS from souce, then use it in HassIO.
Sadly I’m no docker wizard. But your setup sounds like my old one. Running home assistant (not hass.io) on raspbian. In this case picotts runs without any problems. But as I did (the mistake?) and moved to hass.io I’m sitting now in my walled garden and local control and privacy gets harder and harder…
Sad to see actually as in the past @balloob always said local control has kind of priority… I guess these days are over…
As I don’t want any fancy cloud (other people computers) controlled voice recognition and things but just a simple plain (old style?) tts to get some information spoken. Looks like this is a 
only feature in this modern days
I actually did the opposed, I went from HassIO for a VENV based setup.
With HassIO I was getting odd network issues and checking configs and restarting would take forever and HassIO being so locked down made it hard to debug. On VENV things are better for me at least.
I think for some HassIO is a great option especially as it’s easier to set up than VENV, but if you need more control, or want to run a local/offline setup VENV is probably a better option.
One way to address this, perhaps, is via DNS. If you run a local DNS server you can override the “real” IP of the NTP server(s) used in the config to your local machine. Advertise that for just your Hass machine.
You may even be able to put the names of the servers in /etc/hosts with whatever NTP server you want.
Regards,
Ambi
I know this is 2 months later, but they are working on making it completely local:
Maybe not be fully there at this point in time, but it’s headed that way!
BTW, did you figure out how to configure NTP with hass.io? It looks like it should be possible by creating a timesyncd.conf file somewhere, although there is confusion on exactly where.
I think, based on operating-system/Documentation/configuration.md at 712dcf9f74497eef6a75ed1e76f8ae8817ae3ee6 · home-assistant/operating-system · GitHub, and confirmed by OS: allow set NTP by pvizeli · Pull Request #252 · home-assistant/operating-system · GitHub, that it might need to be /etc/systemd/timesyncd.conf
1 Like
What I don’t understand why picotts doesn’t get fixed on hassio. It’s a offline tts which is available for years and works. I used it since ever before I moved from a home assistant pip install to hassio.
So it’s not right that their is no offline tts available at the moment like mentioned from @balloob in the video. It’s just broken at the moment. Their is even a python version of picotts which looks like perfectly for hass/io…
1 Like
Has anyone figured out how to change NTP servers?
My internet is very fast, but not always reliable (very rural cable, sometimes high latency in the 1000’s of mS or intermittent packet loss - in fact had an hour or so outage just today) so I bought a Stratum-1 NTP server appliance that syncs to GPS and I have everything on my network point at that, or at my router (which points to my local NTP server, with NIST as a backup). This has brought most of my systems down from being +/- >1 second to +/- <25mS accuracy (a few experimental things I do actually require that precise of a clock).
I’ve tried to follow directions at https://github.com/home-assistant/operating-system/blob/a592fc986610fac9dbaec96f514ee616c1a44d17/Documentation/configuration.md bit it says the filesystem /etc/systemd/ is read-only so I can’t do anything to change it.
Seems like this should be trivial to add as a setting to the web-UI configuration along with specifying location and units.
I gather NTP is required, and using a local NTP server would guarantee it’s always available, even when internet goes out .
I’m not paranoid about using a public NTP server from a privacy perspective, but I’d prefer to use one that is local and I can have some say in the reliability of so my network continues to function when I have gaps in connectivity to “the cloud”.
hey ,
i didn’t get the last point you mentioned; could you explain more, please ?
what kind of trade-offs ?
Well, if you don’t have internet access the obvious answer is not being able to use anything that requires the internet to function. Like the popular voice assistants like Google Assistant or Alexa.
Then there is the lack of access to any cloud services. For example, ones that I use regularly are weather forecasts and alerts. And of course the recently added coronavirus integration.
Or any cloud based integrations for smart devices (none that I use).
I can understand and agree with the need for the internet for things like updating software, getting the news, weather, and alerts.
However, why should I have to go out to the internet just to get the curtains to open?
Worse still, why would anybody want to make use of any of the “free” services provided by privacy violators like Facebook, Google or Amazon?
Sure, the internet is full of great things and possibilities and people, but it also has more than it’s fair share of baddies who would wreck your environment, DDoS it, co-opt it into zombie farms or ransomware your backups just to name a few things.
We are all so careful to stay at home at the moment and are trying really hard to not get COVID-19, but ok to expose our data to potential abuse. We are so careful against catching a real virus, but are just so happy to catch a ride in the voice assistant limousine where all surfaces are covered layers thick in all sorts of computer viruses, malware and spyware.
More shame us all!
1 Like
I completely agree with you. That’s the reason to use HA and make things work without the cloud if at all possible.
As far as the other concerns, I see your point there too. But each person has to act on their own level of comfort.
However to compare the threat to your physical well-being from a real world virus to the “danger” to you of your computer being “infected” from a computer “virus” or some of your data being stolen is a bit far.
One could literally kill you. While the other is at most potentially financially harmful and usually no more than a big inconvenience or just embarrassing.
Ah, wouldn’t it be nice to go back to the good ol’ days when all you really had to worry about is someone stealing info on your grocery buying habits (first world problems…)? 
1 Like
And for people who don’t care about voice assistants or weather?
I’m unsure if sunrise/sunset requires Internet but that’s the only thing I use that might. For what I care most about it’s looking at stuff on my LAN and at sensors, acting like a really fancy plug-in wall timer.
Also sometimes even if you are fine with Internet connectivity, you may have connectivity interruptions - I run my own GPS sync’d NTP server because I get better precision and uptime of the NTP server than my ISP’s stability provides to the various Internet servers.
About using HA completely offline. I have a sailboat I live on for about 2 to 3 months a year. After HA’ing my house, I recently started doing the same with my boat. After all, it’s pretty much a small floating appartment. When at sea, I only have internet connectivity over Iridium satellites, which is both very slow and very expensive. So I only use it for essential things (like getting weather maps), but certainly not for updating software packages.
I ended up using HA core in a venv on Raspbian. The same setup as in my earthbound home. I get NTP over GPS. The only problem I still have now is that the Lovelace frontend, which can be accessed from small tablets in the cabins of the boat, will try to get fonts from Google. Looks like the culprit is charts.js used for history.
Thanks yeah, but I’d rather not rely on browser plugins for things like that.
Looks like in HA version 0.108.x there’s no request to Google anymore when loading the history page. I previously used 0.103. Seems to be loading fine now when requesting the page on a LAN without internet access. Haven’t updated to 0.110 yet, but unless they added a CDN access back in, HA should work entirely offline now.
1 Like
Hello, does anyone know if this is because the pi doesn’t have a battery to keep a local clock running? I live in a rural area and loosing internet connection and power is common, so if there’s no internet and there’s a power cut, home assistant won’t boot again and I end up with my house lights dead (almost all my lights are controlled with hass) even though power is back. I’m asking because I was thinking that instead of having a local NTP server, a solution could be upgrading to a NUC which should be able to “remember” the time even after a power cut. Anyone knows if that’ll work or I’m better off with a local NTP server? I’m a bit worried about using the local NTP approach since I’d probably go with a pi for that server and I don’t want to end up in a situation were the local NTP won’t start after a power cut when there’s no internet