Mycroft's Goals for 19.08

Originally published at:

We’re now a month removed from the 19.02 Mycroft release. It is so far looking like a great success, so it is time to look out to the next release and talk goals for the Mycroft software at 19.08.

These goals are generally putting deliverables to points from our existing Roadmap(s). They will provide measurable outcomes to guide the development team for the next six months. They include some looks at exciting additions to Mycroft we’re making in the coming months, like deployment of the Persona engine and more focus on Internationalization.

Think we missed something? Let us know below.

Core Tech – Speech Synthesis

The speech synthesis should be of sufficient quality that it is not a distraction to end users. Deploy at least one Mimic II voice using new female recordings.

Mimic 2 was a game changer for Mycroft, making cutting-edge voices possible on Mycroft. Our trial voice was a good start but was also a learning experience in data collection and training. For 19.08, we’ll deploy at least one additional voice using a dataset we’re collecting right now. This voice will be up to industry standards.

Front End – Picroft

Develop a setup experience that a reasonably technical user can engage with to successfully set up a Raspberry Pi.

Picroft continues to be one of the strongest entry points for new Mycroft users. We want to continue to connect with the maker and hobbyist Communities and make getting started with Mycroft easy for them.

Front End – Mark II Setup

Develop a setup experience that a non-technical user can engage with to successfully set up a Mycroft Mark II. This may need to be completed on a prototype.

Mark II’s timeline is not solid, but development continues. In order to ensure less technical people can use Mycroft, we’ll focus on a setup experience that suits them. As timelines for the revised Mark II design are uncertain, this may need to be completed on a prototype.

Front End – Latency

Reduce interaction latency to within 20% of industry leaders.

Making Mycroft faster is a huge deal. We have a portion of the business team focused on poring through metrics coming back from Mycroft Core to identify areas to improve Mycroft’s time to response. We’ll work on the identified areas and move Mycroft’s time to response within 20% of our competition.

Back End – DeepSpeech

Automate DeepSpeech verification and submission process.

DeepSpeech completes the true open deployment of a Mycroft assistant and it needs more focus to improve. DeepSpeech has the potential to finalize the value proposition for offline-only users, speed up the interaction for general users, and gives Mycroft the most compelling business-to-business use case. We’ll move the DeepSpeech feedback loop up to par with Precise.

Core Tech – Precise

Improve Precise to one false activation per 2 hours of test audio

Precise is great for ‘Hey Mycroft’ but isn’t quite perfect yet. We’ll be putting focus on the new audio processing that will be deployed first in Mark II, then future products. Success means one false activation per two hours of continuous test audio.

Front End – IoT

Mycroft stands out from other assistants in its user’s proclivity for IoT. Whereas IoT use is in the middle of most use charts for Amazon and Google, Mycroft’s use consistently puts IoT in the top ten and often the top five. We’ll flesh out the CommonIoT framework for skills and apply it to some of the most used and most versatile IoT skills for Mycroft.

Back End – Persona

Bring Persona data collection online for our Adapt and Padatious engine

Like Precise and DeepSpeech, improving NLP and creating personal Mycroft Personas will have to start with a collection and tagging portal. We’ll have a framework set up for collection of missed intents for Opted-In users, and tagging intents to skills or answering subjective questions.

Core Tech – Internationalization

Have the framework in place for native support of language changing and be able to use Mycroft with all core skills in one language other than English.

The response to Mycroft Translate has really been incredible. For 19.08, we’ll have all of the frameworks in place for the Community to fully deploy Mycroft in their language. We’ll confirm this by testing Mycroft in at least one non-English language for 19.08.

Those are our goals for our 19.08 release. But that’s what we think is needed. What do you think? Leave a comment below and let us know where Mycroft should go.


My first thought on seeing this was ‘what can we take away?’ rather than anything missing that might add to development time - yet still not miss any additions that might be of real benefit and not just a feature.
One option to get Mycroft 2 out sooner might be to:
split some of these goals into release ® and a later update (U1)
Having Mycroft 2 out there would also provide user feedback for additions to update 2
which would neatly set up an improvement cycle embedded in a true user community.

R - One english voice
U1 Female English - U2 other language

R - Standard Release
U1 - Picroft

I would love it to see an more easy noob-way to getting started with mycroft.
To growing it is very importand to find easy ways to bring mycroft to the people.

Yes you have picroft but setting it up is not easy and not really cheap.
For me it seems that we spending not enough time to bring the picroft image forward.

In my opinion there should be:

  1. Native Respeaker support by the mycroft-setup-wizard
    Respeaker by seeed is a company who provide some mic-arrays for the raspberry pi and also some boards with a dedicated soc. Nice, cheap and good availible hardware here in europe. The 4-mic-array has a led-ring but it is not so easy to bring it to run. The wizard should help here too. I think it should be easily possible for you to get in contact with seeed and make a little deal with them to provide a kit in your coming shop.

  2. Input is the one thing but output is the other. It is really sad that the picroft image can not handle bluetooth speaker by default. If that would be natively possible it would be easy for the poeple out there to get some result, to see how cool mycroft is.

  3. Monthly stable and unstable picroft images.
    The last stable picroft image is from 2018-09-12. The last unstable from 2018-12-17. That means that the images are a bit outdated. On my testings some pip-depencies are outdated after doing a mycroft update too or they having a conflict. pip is not on the latest version. All of that is not deal braker but it looks better if the installer has no red or orange lines with warning in it. And on the other side if the image comes with pre-setted raspbian updates, the user can use the time for setting the device up.

  4. Another great hardware out there, with a build in mic, a build in speaker and some power is… all those Android devices!
    Yes I know that this is already on your roadmap but I need to mention it here again. I could be really useful if this topic going deeper into the mycroft-dev area to push it forward. At the moment I use an Android device with a custom rom, the F-Droid app store and without any Google-Apps. A way to use mycroft as easy as Google-Assistant would be awesome! Or what is about Fairphone? They have a Google-Free android rom for there devcies, would be nice if they had mycroft as voice assistent on it, or? :wink:

  5. Get in touch with some cn innovativ hardware distributers like Anker.
    In 2017/2018 the company provide a low cost alexa device without bluetooth but with 3,5" jack and a mono speaker. Here is the community thread.
    I dont know about the internal hardware but a contact from your side could be a chance to bring an low-end mycroft device for less money, out to the poeple.

Please dont understand me wrong. I love your work and love it to being a part of it by supporting you with buying the Mark 2. Using opensource tools and using tools wich give the poeple more control over there privacy is taking allways more time. Google is easy and fast to use, amazon is easy and fast to use, Microsoft is easy and fast to use but all of them are getting quite to much data from the user.
Mycroft is a chance for us all to get rid of theese companys. I understand that you need the money from Mark-2 to pay the team and future updates but some deals with other hardware-contributers like matrix, seeed, fairphone and perhaps other companys could be really usefull. The people waiting for there Mark-2 devices and we trust you. Giving us some better tools to fill the time-area between now and the delivery would be awesome. And because not every baker has a Mark-1 :slight_smile: , we need to choose other hardware like the raspberry-pi 3.

I hope I choose the right words.

Greetings by Suisat


So it looks like you are soon to be far enough along to distribute some hardware and software to people willing to be testers. Can you verify this?

And how can a person volunteer to be a tester and what are the expectations?


is it possible that the voice input and voice output be written to a file? I would like to take the commands and read them for other application uses. Maybe even write back to Mycroft to be said. This feature may already be available which case disregard this request.

The logs contain the transcripts. You could extract them from the voice.log and audio log.

Really looking forward to the new voice!

You could make a skill that listen to the messagebus for

And the output that to a file in the format you want.

Hi all, thanks for the great suggestions and ideas.

In terms of testing, for hardware the team are documenting our prototyping of the Mark II and will release schematics once we’ve validated our builds so that anyone with compatible hardware can build their own prototype Mark II as well.

As for the software side, this is always available on Github and anyone interested in alpha testing can get the GUI up and running in KDE Neon. However please note this is a pre-alpha release so is not intended to be used for anything other than testing and development.