Persian STT & TTS Data Collection

This project aims to provide the codes that have been utilized to collect the Farsi ASR Dataset🤗.

Current Resources

1. YouTube Data Collection Notebook

• Crawls specified YouTube channels.
• Downloads high-quality audio along with corresponding manually created subtitle.
• Batches up the downloaded files in tar.gz files and stores them in HuggingFace🤗 repository.

🚀 Contribution

Please feel free to contribute to improve and expand this dataset! Here’s how you can help:

• Run the existing scripts: Execute the Colab notebooks to expand the dataset and submit a pull request.
• Suggest new data sources: Open an issue and mention YouTube channels with high-quality subtitles or other sources of transcribed Persian speech data.
• Improve data processing: Help refine cleaning, filtering, and segmentation methods.