We've updated DeepSpeech Frontend to enable better transcriptions using Mozilla DeepSpeech! From simplifying Voice Activity Detection to changing our weights to make the best use of the new voice model in DeepSpeech 0.5.1, this update significantly improves the accuracy of text generated by DeepSpeech Frontend.

If you want to help Mozilla in their quest to improve open source speech transcription, check out Common Voice, a project to build quality data that trains Mozilla DeepSpeech, Mozilla TTS and more!

The Machine Learning Group at Mozilla Research also blogs about progress on DeepSpeech, Opus and other Machine Learning enabled endeavors.