Dataset · MLCommons
People's Speech: 30k hours of English speech
MLCommons/peoples_speech
Large open ASR corpus: over 30,000 hours of transcribed English speech.
What's inside
The dataset contains over 30,000 hours of transcribed English speech from diverse speakers. Licensed for academic and commercial use under CC-BY-SA and CC-BY 4.0. Positioned as one of the largest open English ASR corpora.
What you can do with it
- Training ASR models like Whisper
- Fine-tune wav2vec2 for English speech
- Evaluating recognition quality on diverse speakers
- Tasks
- Automatic speech recognition
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 46.9K
- Likes
- 306
Models trained on it
- nvidia/nemotron-speech-streaming-en-0.6bAutomatic speech recognition 106.4K
- nvidia/parakeet-ctc-1.1bAutomatic speech recognition 47.2K
- nvidia/canary-qwen-2.5bAutomatic speech recognition 25.1K
- nvidia/parakeet-rnnt-0.6bAutomatic speech recognition 23.2K
- nvidia/parakeet-ctc-0.6bAutomatic speech recognition 21.5K
- nvidia/parakeet-tdt-1.1bAutomatic speech recognition 20.4K
- nvidia/parakeet-rnnt-1.1bAutomatic speech recognition 10K
- nvidia/parakeet-unified-en-0.6bAutomatic speech recognition 7.6K
How to load it
from datasets import load_dataset
ds = load_dataset("MLCommons/peoples_speech")