
Mozilla Formally Discontinues Its DeepSpeech Project (phoronix.com) 10
An anonymous reader shares a report: One of the interesting projects engaged in by Mozilla that directly wasn't related to their web browser efforts was DeepSpeech, an embedded/offline speech-to-text engine. To not much surprise given the lack of activity in recent years, last week they finally and formally discontinued the open-source project.
Mozilla DeepSpeech was a promising speech-to-text engine with great performance for real-time communication even when running on Raspberry Pi SBCs and other low-power systems.
Mozilla DeepSpeech was a promising speech-to-text engine with great performance for real-time communication even when running on Raspberry Pi SBCs and other low-power systems.
Makes sense (Score:2)
I never understood Mozilla's foray into AI.
There's just nothing about Mozilla that suggests to me they are experts on the subject matter and have much to contribute in the area. I could be wrong, but Mozilla is so tightly associated to the web that it just was a hard sell to me that their AI efforts were going to go anywhere from the start.
Re: (Score:2)
I never understood Mozilla's foray into AI.
Probably just corporate me-too stuff.
When blockc-- er I mean, A.I. is hot, everyone starts doing something with it simply so they can be "part of the latest trend", even if the thing has nothing to do with the business's core product.
Well then (Score:3)
Mozilla DeepSpeech was a promising speech-to-text engine with great performance for real-time communication even when running on Raspberry Pi SBCs and other low-power systems.
If it's open source, so promising, and has such great performance, surely someone will fork it?
no longer relevant (Score:4, Informative)
Other people have already done it. Here's one that I installed and ran on a Pi 5, it works great.
https://github.com/moonshine-a... [github.com]
Re: (Score:3)
moonshine only works in English. The now discontinued Mozilla DeepSpeech supported 136 languages.
Re: (Score:2)
Right, but all moonshine would require is models that are trained in those languages.
Next step (Score:2)
Hey editor! (Score:1)
Where`s the article link?
Re: (Score:1)
I see they updated the README for the project on github.
OTOH the last commit was in 2021, so to some degree this isn't even news.
Good move (Score:2)
I think the idea was good and having Mozilla do it as open source was also good. They didn't finish in time and now we have many open alternatives available. The AI space evolves fine without Mozilla, so why should they waste their resources on that? If the way speech integration in Firefox, they can just pick any open weights model and use it.