openai/whisper
A general-purpose speech recognition model trained with large-scale weak supervision on diverse audio, released as weights and Python code. A single Transformer sequence-to-sequence model handles multilingual transcription, speech translation into English and spoken language identification, replacing several stages of a traditional speech pipeline. The code expects Python 3.8 to 3.11 with PyTorch, and models are installed through pip as the openai-whisper package.
Maintenance activity