Dataset · openslr
LibriSpeech: 1000 hours of English speech
openslr/librispeech_asr
About 1000 hours of narrated English audiobooks with text for ASR training and evaluation.
What's inside
A corpus of approximately 1000 hours of 16 kHz English speech narrated from LibriVox audiobooks. The audio is segmented and aligned with text. The data is suitable for automatic-speech-recognition and audio-speaker-identification.
What you can do with it
- Train or fine-tune an ASR model like Whisper
- Check speech recognition quality and speaker identification
- Create LoRA to adapt ASR to audiobook reading
- Tasks
- Automatic speech recognition, Audio classification
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 130.7K
- Likes
- 279
Models trained on it
- arbazsiddiqui/voxlockAudio → audio 2.6K
- Fritter-ASR/FritterAutomatic speech recognition 1.3K
- TigreGotico/wakehubert-tinyEmbeddings 1.2K
- TigreGotico/wakephonehubertEmbeddings 986
- futo-org/asr4all-lAutomatic speech recognition 794
- udgamlabs/ShrutiAutomatic speech recognition 656
- futo-org/asr4all-sAutomatic speech recognition 592
- Hebrish/SabichAutomatic speech recognition 555
How to load it
from datasets import load_dataset
ds = load_dataset("openslr/librispeech_asr")