ESPnet's YODAS3 Speech Dataset with 1M+ Entries Trends on Hugging Face

espnet · hf · 2026-09-28

yodas3, a speech dataset from the ESPnet team, is trending on Hugging Face. It contains 1M–10M entries in parquet format under a CC-BY-3.0 license, covering ASR, audio-to-audio, TTS, and translation tasks across audio, text, and tabular modalities.

Related event: ESPnet Releases YODAS v3, Largest Open Speech Dataset at 1.1M Hours(2 posts)→

Original post →

More from Research

Research channel →