Skip to content

Dataset · PerSets

Over 385 hours of YouTube Persian speech for ASR

PerSets/youtube-persian-asr

Persian audio collection from YouTube videos with unvalidated transcripts for ASR tasks.

What's inside

The dataset contains over 385 hours of Persian audio tracks extracted from various YouTube videos. The attached transcripts are raw and unchecked for accuracy, so custom filtering and splitting are required before model training or evaluation.

What you can do with it

  • Fine-tuning ASR models (e.g., Whisper) for social media Persian speech recognition
  • Expanding training data for multilingual speech recognition systems
Tasks
Automatic speech recognition
Languages
Persian
Downloads this month
778
Likes
9

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("PerSets/youtube-persian-asr")