unfake/fake_voices

该数据集包含使用XTTS模型生成的巴西葡萄牙语深度伪造音频。数据集通过XTTS模型生成，该模型是一个预训练的多语言文本到语音模型，包括葡萄牙语。生成深度伪造音频的数据来源于CETUC Corpus，由Fala Brasil Group提供，包含101位发言者的140小时音频。数据集主要用于葡萄牙语深度伪造检测，支持音频分类、信号处理和统计分析等研究。数据集结构为每个发言者一个.zip文件，包含深度伪造音频及其对应的文本转录。

This dataset contains deepfakes in Brazilian Portuguese created with the XTTS model. The dataset is based on recordings from the CETUC Corpus, featuring a total of 140 hours of audio from 101 speakers. Each .zip file represents one speaker and contains deepfake audio along with their corresponding .txt transcriptions. The dataset is intended for Portuguese deepfake detection, supporting research in areas such as audio classification, signal processing, and statistical analysis.