TTS/V2V Audio Deepfake Dataset

This dataset was created in 2025 and consists of 643 synthetic audio samples generated using the Minimax.io platform. It includes both text-to-speech (TTS) and voice-to-voice (V2V) synthetic speech. Specifically, the dataset contains 603 TTS audios and 40 V2V audios. The material spans two languages—Spanish and English—and includes 302 female voices, with the remaining samples corresponding to male voices.

This dataset enables multiple downstream applications, including:
(1) the development and training of models for synthetic-audio detection,
(2) the external benchmarking of audio-deepfake classification systems, and
(3) the assessment of model robustness by introducing adversarial or signal-level perturbations to the audio samples.