Whisper

Whisper is a neural network developed by OpenAI for automatic speech recognition, released on September 21, 2022. The model was open-sourced alongside a research paper and a public code repository, with a model card available on GitHub to document its intended use and behavior. OpenAI positioned Whisper as approaching human-level robustness and accuracy on English speech recognition, making it well suited to developers who need reliable transcription of spoken English in real-world audio conditions.

Because Whisper is published as an open neural net rather than a closed API, it fits workflows that benefit from local or self-hosted inference, fine-tuning, and integration into custom audio pipelines. The accompanying paper, repository, and model card give practitioners direct access to the architecture, training objectives, and evaluation details, which helps when adapting the system to specific domains or languages beyond English. For current production audio and transcription needs, OpenAI directs users to its updated models documentation, since the 2022 announcement reflects Whisper's original release scope.