YodaLingua-Danish: 21 Hours of Danish Speech from 925 Speakers

YodaLingua-Danish is a speech dataset containing 7,871 audio-transcription pairs totaling 21 hours of Danish speech. It was created by Thomcles and is part of the multilingual YodaLingua collection. The dataset was last updated on Hugging Face in April 2026.

Use Cases

• Train text-to-speech models based on the high-quality, clean audio-text pairs.
• Develop automatic speech recognition systems based on the well-aligned audio-transcription pairs.
• Fine-tune voice synthesis models based on the dataset's inclusion of 925 distinct speakers.
• Benchmark speech processing algorithms based on the Danish-language audio corpus.

Strengths

• Contains 7,871 individual audio clips, providing a substantial number of training examples.
• Offers 21 hours of total audio content for model training.
• Includes audio from 925 distinct speakers, which may support speaker diversity.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Thomcles on Hugging Face.

Collection Method

Part of the multilingual YodaLingua collection; specific gathering method is unknown.

Freshness

Last updated 2026-04-20 16:53:17; freshness should be verified.

Geography

Denmark (implied by language).