I started to build Leon in late 2017 on my free time. There is no big organization behind Leon AI, just me and more with the community later.
At the moment the whole core is being rewritten, many new features are coming up. Please do know that the main strength of Leon is about the core and not the skills yet. Once the official release shipped, with the community we will all focus on building new skills by extending from the core and let Leon makes meaningful things.
Among the upcoming things, new offline STT/TTS solutions will take place.
It's going to take some time to build a respectable personal assistant that respects our privacy, but we are getting there.
If you have any question or just wanna chat, feel free to join us on Discord.
Thanks for your work on this. Maybe it’s time to introduce a paid license to support the project and even offer consulting hours to get professional support?
Thanks for saying. I have some ideas in mind, like creating some courses specific to Leon. In the meantime, it is possible to support via GitHub Sponsors: http://sponsor.getleon.ai
I don't know how well Leon works but Mycroft is not running locally on your own server. It's still a cloud service. I believe it can be run locally but it's very complex and not well supported. At least this was the state last time I looked at it.
For this reason (and their failed Kickstarter) I've never really been into Mycroft. After all it's kinda the same as Alexa/Siri just a slightly more trustworthy party when they say they're not using your data.
I'm just not interested in alternatives until they can run fully locally (not necessarily on device, on a local server or docker is fine)
If this Leon does what it says on the tin it might be just the thing for me.
Most of my 30 something friends have ditched their Alexa's and Google assistant devices because of concerns around security and privacy. The ability to say "set a timer for 30 minutes" is nice, but not enough to invite always on microphones into private spaces
When I used to use Alexa there were a lot of unexpected things that came up and I used it atleast 15+ times a day. Things like what’s the weather like, what time is it, set a timer, unit conversions, turn on/off lights or appliances, when is it going to rain, factual google information, when was someone born, what’s the sports scores, I like trivia so it could ask you questions while you were lounging around etc. and so many more things to be honest. I stoped using it like 3 years ago though, can’t have an always on speaker in my house that sends all its info back to Amazon, no matter what assurances they give me.
I have both Google and Apple smart assistants in and around my home. I estimate 80% of that usage is to check the weather or set a timer. It would probably be great to not depend on some inscrutable listening device to do so.
My main usages for a voice assistant is adding items to the grocery list when I see something is missing in the kitchen, playing music, asking the weather, and timers.
Sailboat - My wife and I want a solution for our sailboat that allows for intelligent vocal control of ship systems. Quickly launching a series of activities with simple verbal commands or receiving verbal updates on conditions would be amazing. Easy things like automated voice capture to text for log books and maintenance would be helpful. "Oil level at 100%, Oil quality good, coolant level 100%, coolant quality good. Schedule oil change at 1500 engine hours." All of this must be done WITHOUT a constant internet connection.
I've tried to self host myfcroft but after tinkering with it for about 3-4 hours I gave up. They provide little to no information for self hosting. Sure, you can use their pre-built docker image but you still have to create an account in their cloud and connect to it. And their privacy policy is not so great imo.
I did figure out how to do it at one point. I think you had to remove some default config from a JSON file to stop it connecting to their cloud. You still had to query Google directly for the STT though
It is self hosted, offline by default, with options to use various ASR and TTS engines, some online, depending on your own privacy, performance or quality choices. It's quite mature and the maintainers are aiming for a 1 0.0 version release. I have been running it as the primary voice interface to my home automation system for years.
As someone else said elsewhere, there are a few assistants around now. Perhaps there is some benefit for sharing of resources too, as all struggle for contributors.
I think this would be awesome to develop a Tesla "skill" for. Since it's self-hosted, I don't have to worry about sharing my Tesla credentials with a 3rd party. I tried to create a quick and dirty Alexa skill to do things like check on the state of charge and lock the car, but didn't want to climb the mountain of storing "secrets" in Alexa/AWS. Keeping them local in Leon sounds much better.
I would also like to use Alexa to control my doors (garage, house, etc.) but making that available to world doesn't sit right with me. But with Leon I don't have to worry about someone hacking my Alexa account, or even sharing my credentials with a 3rd party "skill".
Readit News
At the moment the whole core is being rewritten, many new features are coming up. Please do know that the main strength of Leon is about the core and not the skills yet. Once the official release shipped, with the community we will all focus on building new skills by extending from the core and let Leon makes meaningful things. Among the upcoming things, new offline STT/TTS solutions will take place.
It's going to take some time to build a respectable personal assistant that respects our privacy, but we are getting there.
If you have any question or just wanna chat, feel free to join us on Discord.
For this reason (and their failed Kickstarter) I've never really been into Mycroft. After all it's kinda the same as Alexa/Siri just a slightly more trustworthy party when they say they're not using your data.
I'm just not interested in alternatives until they can run fully locally (not necessarily on device, on a local server or docker is fine)
If this Leon does what it says on the tin it might be just the thing for me.
It allows you to use the Google text to speech API for more natural voices, which is a cloud product.
It is a yaml config to use on device tts instead.
Install is flashing an sd card.
https://news.ycombinator.com/item?id=19152561
That's a lot of friends! Good job.
rhasspy comes to mind, but i believe Mycroft can be self-hosted too.
Deleted Comment
We'll see if he gets anywhere...
It is self hosted, offline by default, with options to use various ASR and TTS engines, some online, depending on your own privacy, performance or quality choices. It's quite mature and the maintainers are aiming for a 1 0.0 version release. I have been running it as the primary voice interface to my home automation system for years.
As someone else said elsewhere, there are a few assistants around now. Perhaps there is some benefit for sharing of resources too, as all struggle for contributors.
I would also like to use Alexa to control my doors (garage, house, etc.) but making that available to world doesn't sit right with me. But with Leon I don't have to worry about someone hacking my Alexa account, or even sharing my credentials with a 3rd party "skill".