YODAS v3 speech dataset announced with 1.1 million hours of audio
A September 28 post says YODAS v3 is available on Hugging Face, with 48 kHz stereo audio, timestamped transcripts and translations, and 100-plus languages.
TLDR
A September 28 post announces YODAS v3 on Hugging Face, claiming 1.1 million hours of audio and calling it the biggest audio dataset ever. It says the dataset is the first at this scale with 48 kHz stereo audio, and lists timestamped transcripts, translations, 100-plus languages and a CC-BY-3.0 license.
Combined views
1.3K
1 Source, first seen 2h ago
