Egyptian Text-Audio Dataset for TTS Model Training

An automated pipeline for collecting Egyptian Arabic text-audio pairs from YouTube videos. The dataset is created by OmarAhmedSobhy and was last updated on 2026-04-25. It uses forced alignment and automatic speech recognition models to process the audio and text.

Use Cases

• Train text-to-speech models based on Egyptian Arabic audio-text pairs.
• Fine-tune forced alignment models based on the described processing pipeline.
• Benchmark automatic speech recognition performance on Egyptian Arabic dialects.
• Develop pronunciation dictionaries for Egyptian Arabic based on aligned audio segments.

Strengths

• The pipeline is automated, suggesting potential for scalable data collection.
• Utilizes specific, named models (mms-300m-1130-forced-aligner, Cohere ASR) for processing.
• Focuses on Egyptian Arabic, a specific dialect not covered by many general TTS datasets.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Row count, file formats, and column-level documentation are absent.
• Data may reflect geographic and source bias inherent to YouTube content.

Provenance

Source

YouTube videos, processed via an automated pipeline.

Collection Method

Automated downloading, audio processing, and forced alignment using specified models.

Time Range

null

Freshness

Last updated 2026-04-25 16:24:38; freshness should be verified.

Geography

Egyptian Arabic dialect focus.