# Large-scale weakly supervised training

Whisper is a pre-trained automatic speech recognition (ASR) and speech translation model, trained on 680,000 hours of labeled data with strong generalization capabilities