Mozilla's DeepSpeech 0.9 Released For Open-Source Speech To Text Engine

[phoronix]

Phoronix: Mozilla's DeepSpeech 0.9 Released For Open-Source Speech To Text Engine

Following this summer's big round of layoffs at Mozilla, the organization's deep learning open-source speech-to-text engine has been among the projects considered at risk. Fortunately, at least for now, DeepSpeech is still moving forward and is up to version 0.9...

http://www.phoronix.com/scan.php?page=news_item&px=DeepSpeech-0.9-Released

[bug77]

So MDN gets the boot, but this doesn't. It's getting increasingly harder to root for Mozilla these days.
And it's sad, after fighting for so long for an open web, seeing it go back to proprietary crap, like AMP and basically a single browser for everybody

[schmidtbag]

So MDN gets the boot, but this doesn't. It's getting increasingly harder to root for Mozilla these days.
And it's sad, after fighting for so long for an open web, seeing it go back to proprietary crap, like AMP and basically a single browser for everybody

I agree. I don't think we ask much of Mozilla but they keep defying what people do ask for and make something nobody asked for.

[_ONH_]

Somehow they need to find a revenue stream, I don’t see where they can find one in the web space. Depending on google default search seems very risky.

Appart of this, I can see that a speech to text engine could be useful for accessing the web without typing, if it works on mobile device it could enable complete new flays to access the web, without integrated touch screens. And no Siri and Alexa are just 3 year olds.
and if you ask me the interface to Computer/Web is still the biggest issue for me. There is Stil no appropriate solution.

[dnebdal]

So MDN gets the boot, but this doesn't. It's getting increasingly harder to root for Mozilla these days.
And it's sad, after fighting for so long for an open web, seeing it go back to proprietary crap, like AMP and basically a single browser for everybody

IIRC, this project is supported by EU funds and done in collaboration with multiple universities, so they're both obligated to finish it and get paid for doing so - and much of the lifting is done by PhD students, so it doesn't even take that much of their time.

[ix900]

Isn't there anyway to use this without building? Seems like they need to work on ease of use.

[Caffarius]

I agree. I don't think we ask much of Mozilla but they keep defying what people do ask for and make something nobody asked for.

I've been asking for this for awhile. I haven't had time to finish the project but I was having a lot of fun tinkering with Mycroft before I hit the brick wall of having to send voice data to a third party service. DeepSpeech is the last piece we need to host an in-house open source Alexa/Siri replacement.

[bug77]

IIRC, this project is supported by EU funds and done in collaboration with multiple universities, so they're both obligated to finish it and get paid for doing so - and much of the lifting is done by PhD students, so it doesn't even take that much of their time.

Still doesn't yield something people want.

Also, without Siri or Android Assistant proving samples, good luck training it.

[Jaxad0127]

Also, without Siri or Android Assistant proving samples, good luck training it.

That's what CommonVoice is for. https://commonvoice.mozilla.org/

[jntesteves]

Let's not forget Wine is in great need of speech recognition implementation as more software comes out every day that relies on Cortana online services for speech. Including games like Phamophobia and recent F1 games. Speech recognition software is a must for many use-cases. Also important for accessibility. This project can be very important for Linux as a general purpose OS.