Turkish Text-to-Speech Combined Dataset with ~81,500 Audio Samples

A combined Turkish text-to-speech dataset aggregating seven open-source sources. It contains approximately 81,500 audio samples recorded at 24kHz and is described as SNAC-compliant. The dataset was created by user 'afkfatih' and was last updated on November 30, 2025.

Use Cases

• Training text-to-speech models based on the aggregated ~81,500 Turkish audio samples.
• Benchmarking speech synthesis quality using the combined dataset from multiple sources.
• Fine-tuning voice models for specific Turkish speakers based on the listed contributor datasets.
• Conducting linguistic analysis of Turkish phonetics and prosody using the audio corpus.

Strengths

• Combines seven distinct sources, potentially increasing diversity.
• Contains approximately 81,500 audio samples, providing a substantial corpus.
• Audio is standardized at 24kHz and described as SNAC-compliant.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The exact composition and quality of individual source datasets require manual inspection.
• Row count is an approximate figure (~81,500); precise size is unknown.

Provenance

Source

Combination of seven open-source Turkish TTS datasets from Hugging Face, including Mazlum Kiper, Ahmet Deniz, Nisan Kumru, Derya TTS v2, Derya Karma v3, Khan Academy, and Common Voice 17.

Collection Method

Aggregation and combination of existing datasets.

Time Range

null

Freshness

Last updated 2025-11-30 22:10:16; freshness should be verified.

Geography

null