AUDETER: A Large-scale Dataset for Deepfake Audio Detection

AUDETER (AUdio DEepfake TEst Range) is a large-scale dataset for deepfake audio detection. It consists of over 4,500 hours of synthetic audio generated by 11 recent text-to-speech models and 10 vocoders. The dataset was created by author wqz995 and was last updated on April 4, 2026.

Use Cases

• Training deepfake audio detection models based on a large volume of synthetic audio.
• Benchmarking detection algorithms against a diverse range of TTS and vocoder patterns.
• Researching the generalization of audio forensic models across different synthesis techniques.

Strengths

• Contains over 4,500 hours of audio data.
• Includes audio generated by 11 TTS models and 10 vocoders, suggesting diversity.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and exact file formats are unknown, which may limit suitability assessment.

Provenance

Source

wqz995

Collection Method

Likely contains synthetic audio generated by specified TTS models and vocoders.

Freshness

Last updated 2026-04-04 06:15:04; freshness should be verified.