Whisper

Whisper, an AI-powered speech recognition tool, employs large-scale weak supervision and exhibits strong performance. This versatile model can carry out tasks such as multilingual speech recognition, speech translation, and spoken language identification. It utilizes a sequence-to-sequence model that enables simultaneous representation and prediction decoding of sequence tokens. With its five different model sizes, users can choose between varying speed and accuracy levels. Furthermore, Whisper is open-source and operates under the MIT license.