DeepSpeech, the organization’s deep learning open source voice-to-text engine, has become one of the projects considered risky after a round of big layoffs at Mozilla this summer. Fortunately, deepSpeech, at least for now, is still moving forward and has reached version 0.9.
DeepSpeech’s future has been in doubt, even though the project is close to a stable version of 1.0. Although in August they said DeepSpeech 1.0 would be released “soon,” we see that hasn’t happened yet, with the 0.9 release today.
DeepSpeech 0.9 is the latest version of this multilingual, deep-learning MPL 2.0 authorized speech-to-text engine. This release fixes some bugs, adds support for ElectronJS 9.2, improves the score of hot words during decoding, provides optional support for Layer Norm during training, and other improvements.
For more information about DeepSpeech and cross-platform program downloads, please visit GitHub’s DeepSpeech 0.9 details page.