OpenAI Open-Sources Whisper

OpenAI has open-sourced Whisper, its automatic speech recognition technology for transciption and translations.

In a posting on GitHub, where several versions of the Whisper software can be downloaded for free, OpenAI stressed that Whisper is intended for use by "AI researchers studying robustness, generalization, capabilities, biases and constraints of the current model."

It does note, however, that Whisper could also be "quite useful as an automatic speech recognition solution for developers, especially for English speech recognition" and for "certain tasks, like voice activity detection, speaker classification, or speaker diarization, but have not been robustly evaluated in these areas."

Whisper was trained using 680,000 hours of multilingual and multitask data collected from the web, which OpenAI says has led to improved recognition of unique accents, background noise, and technical jargon.