Mozilla Releases DeepSpeech 0.7 As Their Great Speech-To-Text Engine

One of the lesser known Mozilla software efforts is DeepSpeech as a speech-to-text engine built atop TensorFlow with CPU and GPU (CUDA) acceleration. Friday marked a new release of this DeepSpeech software that is yielding great results for converting spoken audio streams to text.

DeepSpeech 0.7 is the new release from Mozilla for this open-source speech-to-text engine. Among the many changes to find with this update are changes around their TensorFlow training code, support for TypeScript, multi-stream .NET support, a new format is available for training data that should be faster, support for transfer learning, ElectronJS 8.0 support, and numerous other changes.

More details on DeepSpeech 0.7 as well as reference binary downloads for all major platforms can be found via Mozilla's GitHub.

For several months now I have been using DeepSpeech as a benchmark to great success and am quite pleased with its capabilities so far as a speech-to-text open-source software solution.