openwhisper

Streaming ASR Dataset This dataset is designed for training real-time (streaming) ASR models, with a focus on handling chunk-based audio processing. It contains standardized audio segments from LibriSpeech dev-clean, processed for streaming ASR applications. Dataset Description Dataset Summary Source: LibriSpeech dev-clean Total chunks: 2,703 Total duration: ~20 hours (1,212.26 seconds) Unique speakers: 40 Audio format: 16 kHz mono WAV Language: English… See the full description on the dataset page:

Source: Hugging Face Hub (orgh0/openwhisper). Metadata imported from the dataset’s Hub tags.