Skip to content
Fan Zeng (Frank)
Back to resources

Libri-Light

Libri-Light provides English audiobook speech for recognition with little or no labeled training data.

Visit official source (opens in a new tab)

Research use

Use its unlabeled partitions and limited-label subsets to study the effect of supervision volume.

Limitations

Recordings come from LibriVox. Check overlap before pairing this corpus with related audiobook benchmarks.

License and access

Dataset license
Unconfirmed for this release
Access status
Direct access

Use the publisher’s resource page and check the terms for the files and version you choose.

Additional sources