1DamnAudio: 221,929 Urdu Audio-Text Pairs for Speech Synthesis

1DamnAudio is a large-scale Urdu text-to-speech dataset created by mahwizzzz and last updated on 2026-03-18. It contains 221,929 audio-text pairs totaling 208.5 hours of speech. The audio files are 16kHz WAV format with an average duration of 3.38 seconds.

Use Cases

• Train Urdu text-to-speech models based on the large collection of audio-text pairs.
• Benchmark speech synthesis quality based on the dataset's high average confidence score of 0.9767.
• Analyze speech duration distributions for model optimization based on the provided percentile statistics.
• Develop multilingual TTS systems by incorporating this Urdu-specific speech corpus.

Strengths

• Large scale with 221,929 samples providing substantial training data.
• Well-documented audio statistics, including mean duration (3.38 sec), standard deviation (1.12 sec), and a detailed duration distribution.
• High average confidence score of 0.9767 suggests good alignment quality.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset's source, collection method, and potential speaker demographics are not described, which may obscure bias.

Provenance

Source

mahwizzzz on Hugging Face

Collection Method

Method of gathering is not described.

Time Range

Temporal coverage is not described.

Freshness

Last updated 2026-03-18 06:12:09; freshness should be verified.

Geography

Spatial coverage is not described.