Originally published at: Mycroft Manufacturing and Product Update - Mycroft
Mark II Update
Another big month, and some big hardware and software updates to share with you all. In short, we’re moving fast and on track for shipping the first production run no later than September 19. At this time we are taking pre-orders for our third production run (shipping in November).
We are making the final touches on the housing design, such as bosses, ribs and final fit with the various custom and purchased components. We will be 3D printing samples to confirm final fit, followed by engineering tests to ensure it meets our drop and heat test requirements. It will then go for FCC electromagnetic testing to ensure it does not cause interference with other devices. At that time, we will also submit the final design to our vendor to create the injection molds that will be used to create the plastic housing.
We’ve made some aesthetic improvements to the housing of the display and camera. The specifications will not materially change.
We have run into an issue with obtaining the speakers; the vendor pushed out their delivery date. We are reasonably confident that we can obtain a good alternate speaker.
There have been some layout changes to the SJ201 audio processor board. These were to reduce chances of electromagnetic interference, improve voice activation detection, and some other minor changes to improve the user experience. The changes are completely compatible with previous revisions of the board and no software changes are required for this, so existing Dev Kits can run exactly the same software as final production units. One significant improvement from the developer perspective is that we can now reflash the ATTiny firmware from the Raspberry Pi. On the downside, we were unable to find a replacement for one component that is no longer in production so we’ve had to eliminate the external headphone jack. However, for those who really need one, the Mark II now sports three free USB ports and can easily accommodate an inexpensive audio dongle.
We will receive the first lot of these updated SJ201 boards in late April. We will use these in our first New Product Introduction (NPI) testing, which verifies our manufacturing processes through Engineering Validation Testing, Design Validation Testing and Production Validation Testing.
Lastly, we worked with our contract manufacturer to produce several prototype Mark II units. We are close to having all parts in our inventory or on order. We continue to be confident in our commitment to ship no later than September 19.
The Mark II software is nearing completion and the development team has progressed to polishing the user experience. Some highlights include:
- Improved intent matching across the core Skills.
- Caching the locally generated common and expected dialog making most interactions even quicker.
- Scrolling long strings of text when they don’t fit on the screen – such as artist names, song titles, or even Wikipedia page titles.
- Improving the mechanisms that the Common Query system uses to determine the best answer to a given question.
- Expanding the wi-fi connection process to allow you to connect to networks with captive portals, e.g. hotel networks.
- Further improvements to the boot and pairing sequence – making it faster and more intuitive.
- We turned off the internal LEDs… because you couldn’t see them except at night when you don’t want to!
- The display will now dim appropriately when the device is idle. This makes it a lot nicer to use at night or in the dark, reduces power consumption and improves the life-expectancy of the display.
Despite our focus on making the Mark 2 software production-ready we have also developed a new state-of-the-art Text-To-Speech (TTS) engine: Mimic 3. This versatile system is a vast improvement over both Mimic 1 for offline TTS rendering and Mimic 2 for cloud-based TTS. If you have seen any of our recent videos you might have noticed that Mycroft is sounding better than ever- and that voice is now being generated on the Mark II itself!
Moreover, Mimic 3 already has more than 100 unique voices and can speak over a dozen languages! We’ll discuss the details and even more exciting features in future updates.
With the Mark II hardware and software approaching a production-ready state, we are now shifting our focus to our cloud services. These include device and fleet management, anonymous access to third-party services, community collaboration tools and higher quality Mimic 3 voices. More on that next month!