Turns out I was wrong…
21MB… Lol
https ://9to5google.com/2016/03/11/google-accurate-offline-voice-recognition/
1 Like
SAPI voices are ready made voices that can be purchased through companies like CeraProc. They are developed for use with Microsoft COM (component object model) Speech API. Because I am not a developer I have zero concept of the difference between that architecture and what Mycroft is built on, or whether the two are compatible or adaptable. But I already own a few of these voices I purchased from CeraProc, so it would be convenient if I could use those with Mycroft.
Quite impressive. The biggest issue I see is the system they’ve implemented has a fairly limited dictionary. They also run on a nexus 5 for the testing, which has 2x the ram and about 2x the cpu power…which undoubtedly helps. Wonder if they’re using the gpu parts for that as well. Most pi-type boards don’t have similar capabilities yet, but hopefully that changes.
2 Likes
Thanks for pointing me to this @nate-mycroft. This is a very interesting suggestion, I’ve not looked at this system in any detail but it may be possible to integrate in the Mycroft infrastructure despite SAPI beeing closely liked to the Windows platform.
There are some implementations that may be useable, I will continue to investigate.
1 Like
…adjust or mute the volume, operate my window blinds, or my air conditioner, anything with a remote… 
1 Like
@owdha I know we have talked about this, but wanted to document here too. For some reason when I got my echo, I immediately put it in my kitchen, but when I brought home Mark 1, I just instinctively put it in my home office. It just felt right. So I would love to see a form factor that builds on that. Although my desk is probably the messiest one around, I tend to like devices that help me curb the overwhelming demand on my real estate of my desk. So things like built in chargers/USB hub, or even docking type features seem good to me. ( I know I am in the minority probably, but just wanted to put it out there) I would love it to be one of those things that anytime someone comes to my desk they ask me about it, or try and talk to it.
I have a few suggestions:
I know it has been mentioned multiple times but better range for a Mic and voice recognition is paramount.
I would love to be able to use it in much larger rooms.
I like the idea of having multiple types or cases.
A larger one with a bigger screen and camera,
A smaller one with indicator lights.
I would suggest a third one that sells for cheaper and that is just the case. Say I already have bought several PIs I would love to be able to buy just the case you all have and integrate existing hardware.
I’m still really new to Myocroft so I can’t make too many other suggestions, due to the fact that you may already do a lot of what I want to see and do with Myocroft. Such as integration with chromecast to play youtube videos. I’ll find out as I play and make more suggestion based on the capabilities I want
1 Like
Thanks for joining the conversation ssanders9012! We are for sure having a bigger screen on V2, sound quality and Mic improvement are one of our Priorities!. Camera and different ranges of products are great ideas, It will allow more freedom in how you integrate Mycroft around your home.
Q: What would you want to use the camera for?
@ssanders9012 @lsimngar @baconator
Video calls, security camera, I’m sure I can find lots of ways to use it. Just like there are a lot of potential ways to use mycroft in general.
1 Like
“what does mt everest look like?”
"Show me a map of Lawrence."
A visual timer would be handy.
Clock…
…stuff…things…
1 Like
Hi! Mycroft was only brought to my attention recently - I really love this idea and want to see it succeed.
But all of the feedback here is telling me that the existing mic doesn’t cut it and to be frank this was my number one concern about the device, as the demos on the buy page looked like pretty small rooms - in addition the actor needed to stop what they were doing - specifically address the device itself in a crystal clear quiet environment.
Because this is open source - this is not necessarily a showstopper! I’m contemplating putting together my own custom one - but I’d prefer not to have to.
As you’re asking for ideas -
I think you should chuck some more capabilities at the device. if it’s to interact with everything, you’d be well aware there’s plenty of communication standards.
I had a quick look online and there’s some really nice raspberry pi shields you could use to extend the device.
First one I found that looked any good was this:
https://creator.matrix.one/#!/index
This ticks the boxes of a decent microphone array and adds a crapton of features to the device which is all about interacting with IoT.
As is, the Mark 1 is a stationary device. Why does it have to not move? why not give it at least some portability?
Add some battery backup to the device - preferably rechargable batteries (I know adding batteries complicates the shipping). Kids can pick up mycroft and take it around the house etc and ask it questions. mycroft can tell them when its disconnected from wifi or other things like batteries running low or if it’s been left in the sun, can make comments about being rough-housed etc.
Even for adults this would be useful - where i’m from, we also get blackouts, and it’d be good for it to work even when there’s a brown out without waiting for reboot, or damage to data etc.
Further thoughts on voice assistants and AI
Your competitors are google home, and the echo. As is, any of these products do not respect my privacy. Google home, Alexa - all of these are basically devices that are designed to slurp up audio data and ship it to the cloud where I don’t control it. I’m not quite sure what Mycroft uses - pocketsphinx is mentioned in the github which is an offline tool, but so is googles API? it’d be great to process this stuff into a command locally, and send that off to cloud services if necessary. privacy is a selling point and it’s not mentioned at all! people seem to want to happily run in and add always on microphones all over their house that send to the cloud - but I’m not a big fan of this.
1 Like
- Motion detection to detect movement and do something…
- face or objects recognition, but I don’t know if the rp3 GPU is power enough to do that in real time or at least in a few seconds. I didn’t try this yet…
Thanks for all the ideas. @posty I like the idea of a portable Mycroft. I’m going to look into the power requirements. It could also be cool to utilize conductive recharging like Wireless Qi. We also have a matrix dev board, respeaker and a few others that we are evaluating to improve the microphone quality.
2 Likes
So I guess Mark 2 is relatively finished? Is there going to be a blog post? Can we get any more details?
I’m not a fan of the design really and without knowing anything about the internals I don’t know how I feel about it yet. The lower mentioned price is good though.
1 Like
@klundry The Mark 2 design is definitely still in progress. That was an early prototype version that we brought to Techcrunch Disrupt to show off what we have been thinking. We are still very interested in your feedback.
We are still working on the internal specs, but I can tell you it will have a multi microphone array with far-field voice capture, noise/reverb/reverb reduction/echo cancellation, a higher quality loudspeaker design, an LCD display, and at least one USB input/output.
Can you give us more detailed feedback about the design? What would you like to see?
Thanks!
1 Like
May be a smaller version of mycroft. Microft mini.
I’ll second klundry’s motion to add an IR receiver and blaster so that Mycroft can act as a universal remote. This would be cheap to add (Flirc HW and open-source SW exists) but would be a big differentiator from Amazon/Google products and would leverage the multiroom capability envisioned.
Example use case: I’m heading downstairs to watch TV. I tell Mark II in the upstairs living room “Mycroft, let’s watch Plex on the TV downstairs.” As I head down, it passes a command to the Mark II in the family room to send power-on commands to the TV, receiver, and nVidia Shield, then switches the receiver to the correct input and sends a remote keypress sequence to the Shield to launch Plex.
Obviously, making this work requires building a web page UI or a companion iOS/android app to let the user record IR commands from OEM remotes (or better yet, download the codes from a user-contributable online database, as Logitech Harmony does), input and save IR output sequences as activities, and assign unique names to the sequences that Mycroft can recognize and pass as parameters. Plus the unit-to-unit communication.
But the necessary components to build this capability inexpensively already exist, and once working the user experience could be pretty magical the capability could be highly extensible, eventually beyond IR to include Bluetooth, Zigbee, etc. (though those involve pesky licensing).
3 Likes
More specifically what I don’t dig about the prototype is the facial animations and the very rounded tube aesthetic. I suppose it is personal preference and you’re not going to be able to please everyone with a single design but I prefer a more boxy enclosure and the more abstract led eyes and led matrix mouth of the Mark I. I think it is good to make digital assistants appear more computer and robot like. I don’t particularly like the attempts to give them faces and more human like features. Maybe it would be possible to combine an lcd and the led matrix and eyes? So it would be the same eyes and matrix as the Mark I with a high res lcd below for displaying more detailed information when required. There is something about the rudimentary and abstract nature of the led rings and matrix animations that I find cool.
@klundry Thanks for sharing your feedback, Mark 2 face is also a work in progress. We are working towards a more neutral/flat style more than realistic/humanized. However, is a feature that will be customizable. We are working on providing the user an alternative and eventually give users the ability to create their own.
Thanks!