More (Local) Self-Contained Internet-Free Voice Response Solutions



 The Snips Platform: "powering (no web) Private-by-Design voice assistants."
is major project with user generated,  ready to use application examples:



Their project announcement 

"Snips runs on a variety of platforms, including Raspbian, Android, iOS, macOS, and most Linux flavours."  (but for a change it looks like Windows may be the one left behind, for now?)

They have partnered with SeedStudio to provide hardware kits for making embedded voice controlled devices, (you add a Raspberry Pi 3B+)

Here's a Curated List of applications.  github.com/snipsco/awesome-snips/

 One example, "overhead" reports the commercial flight passing by when you ask "What's Overhead", it accesses flightradar24.com for live air traffic info.
github.com/hcooper/overhead

Click Images If You Want to Count Planes...









With $13M Series A financing, one wonders what sort of toll one may have to pay to access it, apparently it's free to use for non-commercial applications under the terms of: (GPL V3 and ad-hoc license on demand for special needs) .



-------


Mycroft "the world’s first open source assistant"

Mycroft Hardware Ver. 2

  This is Not a completely self contained voice response interface, it does ship your raw speech to their "Home" server  for speech to text (although there are plans to offer a Personal Server for speech recognition)

 They speak of the better security of their open source product offers, although I am no clear what makes it more secure, because there's no motivation to commercially exploit the data?  Voice generation is done locally, which is I believe is different from Siri/Alexa.
 They may not be fully up front with users as to their for profit open source revenue sources, stock offerings mention an enterprise server license for $1,500/month, and a $2/month premium membership fee.


There's a collection of user submitted "Skills" for special purpose applications.


  They are just finishing their ver.2 Alexa like hardware interface, it uses an LCD to display the time and show animated speaking lips?.
 Their blog does share product development details, including the use of Teksun.us (US/India DFM, Design for Manufacture) to produce rev 2 hardware, including plastic housing,

Ver. 1 hardware demo from 2015 (maybe a faked computer voice)


Mycroft quality is not quite up to par with the big players massive computing power. Here is a voice quality comparison.


Mimic2  text to speech synth, based on the Google Tacotron project concepts, runs underTensorFlow, it's Trained with 16 hours of audio from a single speaker.

Their blog touches interesting Open Source issues, as well as progress reports on their development progress.

Comments

Post a Comment

Popular posts from this blog

LED Driver Roundup

Image
LEDs being current, rather than voltage operated devices require special driver circuits for optimal performance.  While simple applications like an indicator light that might be run at a fraction of it's rated capability, a series current limiting resistor is sufficient.  High power application running at the edge of the LED's max current rating require active regulation to compensate for the negative temperature coefficients of the LED, (measurement techniques are discussed in this university lab paper .) Low power LEDs can be driven with simple series resistor current control when powered from a regulated power supply. The LED Load Line graph be low helps one select a standard resistor value to get the ballpark current setting with a 5 volt power source.  Find where your LED's color line intersects the desired current, then pick the closest resistance line below that intersection. (This graph is from LEDnique.com , a collection of basic articles on LEDs aimed
Read more

Hooking up the 6 + 1 Mic Array to the MaiX BiT

Image
The Risk-V chip on the MaiX BiT board also contains hardware FFT which should make it a respectable audio analysis board.  Sipeed offers a microphone array board with 7 mics and 12 RGB LEDs. Unfortunately the mic array, which SeeedStudio  indicates is for Dock/Go/Bit, there is no real provisions made for the newer BiT board. Included with the mics is a 10 pin ribbon cable for the Go board, and an unadvertised (not on sales page inventory) 5x2 "dupont" connector cable which directly connects between the back of the mic array (not pictured on the sales page) and the Dock board area with mic connection labels silk screened on the board. Some improvising is needed to hook up the BiT.  I offer 2 different solutions. The first that Seeed could help by throwing in a couple of replacement connector housings. Connection Method #1 (You might have to start with a larger housing and cut it down to size) The Seed OPL type for thes
Read more

Holographic Fans Getting Real Time HDMI Input

Image
 One of the technical leaders in so called 3D Holoographic LED Fans is Dseelab, in Sep 2019 they announced at an event the development of real time HDMI Video Input, it may be quite a while befiue it reaches the market.  Achieving this has been my holy grail, looking into FPGA based video processors and examining cartesian to polar code may no longer be necessary... Here's an interesting demo recently posted: Dseelab a leader in the "3D Holographic LED Fan" industry, has master-slave connections on their units, allowing them to be syncronized in an array, they also offer a programming API connecting though WiFi or with a wired serial connection to a PC.  The API supports remote update from the cloud and random play of 256 different media assets. The only other Holographic Fan maker offering a programming API and arrays for large image display is HYPERVSN of  Ireland.  The newest addition to the DseeLab line is their: 6 Blade 78cm, 960*960 res, 84W Model 
Read more