Skip to content

Dataset · espnet

YODAS-Granary European speech corpus for ASR and translation

espnet/yodas-granary

A large multilingual audio dataset (10M-100M entries) with pseudo-labels for European speech recognition and translation.

What's inside

YODAS-Granary is a filtered subset of the larger nvidia/Granary dataset published by the ESPnet project. It contains 10 to 100 million audio-text pairs with high-quality pseudo-labels for automatic speech recognition (ASR) and speech translation (AST) tasks across European languages (including English, German, Bulgarian, Czech, Danish, Greek, and others).

What you can do with it

  • Training and fine-tuning multilingual automatic speech recognition (ASR) models
  • Training direct speech translation systems for European languages (AST)
Tasks
Automatic speech recognition, Translation
Languages
Bulgarian, Czech, Danish, German, Greek, English, Spanish, Estonian and 4 more
Size
10M–100M rows
Downloads this month
35.1K
Likes
43

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("espnet/yodas-granary")