Emilia-YODAS: English Audio Samples Compressed with NeuCodec

Over 30 million English audio samples totaling more than 78,000 hours comprise this dataset. It is derived from the English-language subset of the Emilia-YODAS collection and has been compressed using the NeuCodec model. The dataset was created by neuphonic and last updated on August 14, 2025.

Use Cases

• Training neural audio codec models based on the NeuCodec-compressed audio data.
• Benchmarking speech synthesis systems based on the large-scale, English-language audio collection.
• Developing speech enhancement or audio restoration models based on the compressed audio samples.
• Researching audio compression techniques using a dataset specifically processed with NeuCodec.

Strengths

• Contains over 30 million individual audio samples.
• Represents more than 78,000 hours of English speech audio.
• Specifically processed with the NeuCodec compression model, which suggests a focus on modern neural audio codec applications.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Freshness should be verified as the last update was on 2025-08-14.

Provenance

Source

English-language subset of the Emilia-YODAS dataset.

Collection Method

Audio samples compressed with NeuCodec.

Freshness

Last updated 2025-08-14 10:48:01.