krisgesling

If you download Mycroft today, currently by default everything except the STT is running on the Pi. There are options to do this locally: https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customiz... however they have their limitations.

Before the Mark II ships in September, local STT will be available by default!

On device STT will be available before the Mark II ships in September!

There are some options for that available if you're running Mycroft already: https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customiz...

But these are not what will be shipped on the Mark II...

Thanks :)

I'm going to assume this isn't our developer Mike's alternate account congratulating himself lol

It's all a question of what data we have at the moment.

If you know of any good Portuguese voice datasets we'd love to train one. At some point we'll start getting new professionally recorded data for each language too.

Haha yeah Thorsten the person behind the voice is such a nice guy - I'm not sure that "angry" is really in his nature :D

No Android release yet unfortunately, but you can drop your email in the bottom of this page and select the platforms you are interested in to be notified about specific future releases: https://mycroft.ai/mimic-3/

Yeah we aren't using the seeed array in the final Mark II. But we have used the same XMOS XVF-3510 to perform acoustic echo cancellation. That means, even with music blasting out of the speakers, you can still wake the device from across the room.

In a simple fashion you can think of it as subtracting the audio being output from the audio coming in from the microphone.

On device STT will be available before the Mark II ships in September!

On device STT will be available before the Mark II ships in September!

Note: I'm from Mycroft

On device STT will be available before the Mark II ships in September!

As will a new privacy policy that better reflects what we actually do.

Translation: "Yes"

... Hi from Darwin :D

Bloody oath we do!

That's some good feedback thanks!

The other option is to jump on https://mycroft.ai/mimic-3/ and use the "Hear my voices" tool. There are three drop-downs: - language - model - speaker

The Mozilla Common Voice dataset is awesome - however it's useful the opposite purpose - speech-to-text. This is because it is a lot of different people using a range of hardware, speaking similar phrases.

For good text-to-speech you need 1 person speaking different phrases but very consistently. Here's an example dataset from Thorsten a German open voice enthusiast: https://openslr.org/95/

We do get this request a bit but not yet at the scale where it is economically viable for us to do so.

I would however point out that the Mark II is completely hackable. So whilst it's not an SJ201 on its own, you can absolutely pull the whole thing apart, use it in other enclosures and even put it all back together again.

Another one of the reasons we made it is because having a single daughterboard greatly simplified production and made the Mark II more robust overall. There's no longer the possibility of a loose wire to the power supply or amp after it gets kicked around in the back of a delivery van. They're all in one and connected via the 40 pin GPIO header, but absolutely removable from the Mark II unit itself.

The audio for all of the voices produced from public data sets is licensed as Creative Commons Attribution Share Alike 4.0 International. This is what a lot of the original data is licensed as.

Full license text here: https://github.com/MycroftAI/mimic3-voices/blob/master/LICEN...

and an explainer on what the legal mumbo jumbo means here: https://creativecommons.org/licenses/by-sa/4.0/

Hey there, instructions for installation are all here: https://mycroft-ai.gitbook.io/docs/mycroft-technologies/mimi...

and the source code is available at: https://github.com/MycroftAI/mimic3

There is currently a Docker image, DEB package, and PyPI release.

Hey there - the source for Mimic 3 can be found here: https://github.com/mycroftAI/mimic3 It can run CPU only or accelerated with a GPU.

Regarding the BOM, I assume you mean the Mark II? That you can find here: https://github.com/MycroftAI/hardware-mycroft-mark-II/tree/m... We actually ended up designing our own RPi daughterboard called the SJ201. It's mostly an audio front end with an XMOS XVF-3510 and dual mics, but also includes a 23W amp, some LEDs for feedback, buttons, a hardware mic switch, GPIO breakout and power management (amongst other things).

Yeah I think there are two sides to this coin (and just for clarity - all of this relates to Picroft, not Mimic 3 the TTS engine that just launched). The audio hardware makes a huge difference to audio input which is why we've developed the custom SJ201 board that's in the Mark II. But even on DIY units we have been making big improvements on the wake word detection by better balancing our training data sets. Once the Mark II is shipping there are additional wake word improvements on the roadmap. Eventually the system will optimize for the users of each device. So the wake word model on your device wouldn't be exactly the same as the model on mine. We've also ported the Wake Word model to Tensorflow Lite which means it uses a small fraction of the system resources that it used to :D We're also about to make some bigger changes to mycroft-core that will help to support a broader range of hardware in a more consistent way. So whilst you could try it again today and I can guarantee it's better than the last time you used it, if you want a DIY system instead of a Mark II - I'd suggest adding a reminder to check it again in a couple of months once these bigger changes land.