Mozilla's Incredible Speech-To-Text Engine Is At Risk Following Layoffs
Reuben Morais of Mozilla provided some guidance today on the state of DeepSpeech following the recent Mozilla changes... But the state right now is basically more uncertainty. There is no clear guidance right now if DeepSpeech "will have a new home" or not at the restructured Mozilla or what changes may be involved.
That's quite a pity as Mozilla DeepSpeech is among the best speech-to-text engines (if not the best, certainly the best among open-source options) that supports real-time translation on a wide range of hardware. DeepSpeech utilizes deep learning based on Baidu's research and leverages Google's TensorFlow. DeepSpeech is quite a quality piece of software and has delivered excellent speech-to-text results for translating audio into accurate text. I've personally experimented with it a lot as part of DeepSpeech benchmarking in evaluating its CPU performance.
Making matters worse is that DeepSpeech has been approaching its big "v1.0" release as their first official stable debut. Reuben at least indicated that since much of the 1.0 technical work is already done, they are at least planning to ship DeepSpeech 1.0 soon. "Most of the technical changes were already landed, and we see no reason not to ship it. We’ll be releasing 1.0 soon and encourage everyone to update their applications."
But following the 1.0 release, there is no certain future for this speech-to-text engine at least under the Mozilla umbrella. The developers though do encourage continuing to submit bug reports, pull requests, etc. The brief statements were made today on the Mozilla Discourse. If Mozilla does away with DeepSpeech, at least hopefully some other organization will decide to batch this STT engine under its open-source license (MPL 2.0).
Those wishing to learn more or experiment with DeepSpeech can find it over on GitHub. The releases page does have DeepSpeech binaries for all major platforms.