Whisper

Whisper is a powerful artificial intelligence-based system designed for speech recognition. It utilizes a large-scale weak supervision setup and is a comprehensive model suitable for multi-lingual speech recognition, language translation and spoken language identification. It works through a sequence-to-sequence approach that combines sequence tokens and predicts decoding. It is open-source and comes with five distinct model sizes that can be adjusted for desired speed and accuracy. All of this is provided under the MIT license.