I don’t agree with that assessment. None of the options below are perfect and have caveats, but there are local options. In addition to Sonos as already mentioned above.
That would be a waste of their time and our money, IMO. The hub would still need to call out to AWS for the initial TTS file. And then the announcement would be played over a small, tinny-sounding speaker built into the hub? No thank you…
Please no. I don't want to pay for more stuff stuffed into the hub (e.g., others are also asking for POE support) that I dont' want/need integrated. Also just plain don't want the size of the hub increased, I like the hub as small and unobtrusive as possible. The antennas on the C8 are the exception (don't mind the additional "bulk") as they enhance the core functionality of the hub, controlling devices/running automations.
We have our hubs in the office, best place for them to be close to the middle of the house. If they were in there announcing stuff all day my wife (who works in there a lot) would kill me and then burn the hubs. 
Everyone in this forum probably has one or two "If only the hub also had this built in" dreams, but if you go down that rabbit hole you end up w/creeping (creepy? 
) feature clutter and physical device bloating that turns a cool sports car into: 
I guess we disagree on this. We have 500 models of contact sensors yet no off the shelf option for tts. There is a clear need for a device.
I think the main disagree is about where that TTS speaker physically lives...many of us don't want it in the hub. 
I can certainly understand the appeal of a fully local TTS solution. Just not one that’s fully built into the hub itself.
We used to try to get cheap Fabriq speakers to work on HE ...they never really did well. I'd like a non-Sonos HE speaker for a few spots.
I have done two options.
The Google Nest Mini's work well for the most part. You can disable the mic with a physical switch instead of the app. Not sure if that really makes a difference though.
The VLCThing was a really interesting experiment. Once I got the steps down it wasn't to hard to setup, but figuring it out initially was a pain. It actually worked well except for the occasional connectivity issues with Bluetooth. It simply required a raspberry Pi w (computer) and then the speaker. The pi could be replaced with any computer device. The magic sauce is VLC being configured properly and VLCThing on HE. If i was to do it again I would use a wired speaker device to remove the connectivity issue.
That's back in the day. I have my Fabriq around the house somewhere...
I don't want the speakers (for the solution I'm looking for) to be in the hub either. I also don't really want the speakers to be "smart". I don't think speakers have to be "smart" to be "wireless".
I've read up on the local TTS solutions mentioned, and they don't seem to be ready for prime time. Most haven't had any posts/comments in quite a while. I'm not downplaying the work done by others on those solutions, but I don't think any of them are near the point where, for example, the HE developers would consider them to be equivalent in reliability, functionality, etc. to the built-in apps.
I'm very new to the HE environment, and part of my problem is that I always want to do something similarly to the way I did it before. So as previously mentioned, my previous setup was an Elk M1G. The speakers were as dumb as they can be, currently priced at about $11 each. I had several of them connected/hardwired to the M1G, and they all sounded "in unison". They were speakers, nothing else. The M1G had a 500 word vocabulary from which an announcement could be "built". It also supported "custom prerecorded messages" that could be uploaded to the M1G.
For what I want, I may be looking at this wrong. Maybe the best solution for me is to continue to use the M1G for announcements. Most of the events that will drive announcements will be detected by sensors connected to the M1G, so the M1G can still announce those. However, for various reaons, I plan to use z-wave sensors controlled by HE for other events, e.g., water leak sensors, temp/humidity sensors, etc. The M1G's vocabulary is sufficient for pretty much all of the announcements I would want HE to make, so maybe the best solution for me is to somehow have HE tell the M1G to make a particular announcement.
For example, if upon detecting a water leak, HE could trip a M1G zone that doesn't have a physical sensor connected to it, then a M1G rule could trigger an the appropriate "water leak in bathroom number 1" message over its speakers.
The problem there is it looks like the Elk M1 Integration app isn't being fully supported any longer.
Amazon echo has worked flawlessly for me..obviously not local:
But for my use case wouldn't want it to be.. (ie alerts in multiple houses)
Actually, there is 100% local option available, however it is a little less flexible.
I have an automation that plays an .mp3 file on my IKEA Symfonisk speaker (Sonos) any time mail is delivered to our house. I uploaded an .mp3 file to the Hubitat Hub, and then use Rule Machine to tell the Speaker to play the .mp3 file.
You could create a bunch of audio files, one for each of your TTS requirements. This way, they would all be 100% local.
Please be the AOL “you’ve got mail” sound…
lol, I'm still using them with Node-red and Hubitat for playing mp3 when a door is open of left open, as well as alarm sounds, never had a problem. I also have them blocked on my router so they never dial home. 
There are a couple of issues with that solution (again, because I'm looking for something similar to what I had before)...
I think any solution that is primarily controlled by HE will require too many compromises for me. Same goes for cloud based solutions. I don't think I would be allowed to put Echo Dots or HomePod minis that need power 8' high on walls in hallways. Even if I could, I think there would still be a problem with them not sounding at the same time.
The app itself is a "beta" release, but it was made available for everyone to try with the 2.3.6 firmware update.
Nope! It is an audio clip from Blues Clues. Gives our young adult children a laugh every time they hear it. 
One thing I am willing to compromise on is "100% local". Since HE only needs internet access the first time it sees a unique text, and the cached result will survive a reboot, I am okay with that.
That leaves the issue of requiring multiple wireless smart speakers instead of using dumb wired speakers that are at most 4"x4"x2".
Is there a device (or group of devices) that HE can send a digital "announcement" (in whatever form/format that may take, e.g., wav or mp3 file), and that device converts it to analog and sends it to the speakers? My computers still have a "line out" jack that can drive small speakers with built-in amps. Can this be done?
I'm probably using some terminology incorrectly, but I hope everyone can understand what I'm getting at.
Seems simple enough, but I don't know what all HE or the rPi can do.
Just to make sure it is clear... The behavior described above is how Hubitat's Text To Speech works when using the built-in Sonos integration.
Echo Speaks (community integration) relies on the Amazon cloud for every TTS operation. However, I am pretty sure there is a Broadcast feature that plays the TTS on all Echo devices at the same time, which is nice as the audio is synchronized throughout the house. It has been a long time since I used Echo Speaks, so I am not 100% certain if this is still the case. Also, some Echo devices have a headphone audio output jack. So, this might be a way of generating one analog audio signal that you somehow wire up to a bunch of speakers throughout your house.
The Hubitat built-in Chromecast integration, used with Google Home devices, probably relies on the Google cloud for every TTS operation as well. However, I am not 100% certain about that one.
Hubitat's built-in Apple Air Play integration is relatively new. I am not sure if it takes advantage of Hubitat's caching of TTS requests or not. @gopher.ny from Hubitat could probably clarify this one.
Using VLC running on a Raspberry Pi or other always on computer is another option, as @mavrrick58 and @marktheknife both mentioned earlier. I am not sure if this technique relies on the cloud or not? 
There are definitely a lot of options.
Is VLCThing an ongoing project? I found a lot of posts referencing it but I was unable to find a "custom app" page for it with details on what it is, what it does, where to get it, etc.
Download the Hubitat app
