Voice Cloning and Forgery Detection Using WaveGAN and SpecGAN

This paper presents a comparative analysis of Deep Convolutional GAN (DCGAN), WaveGAN, and SpecGAN for voice cloning and forgery detection. The aim is to assess how well each GAN performs in producing excellent in quality synthetic recordings of speech and identifying artificial voices. Experimental results show that SpecGAN outperforms both DCGAN and WaveGAN in generating high-quality synthetic voice samples. In addition, the paper proposes an audio forgery detection method that combines copy-move forgery detection, CQSS-GA-SVM analysis and audio forgery detection using SpecGAN achieving 98% accuracy in detecting synthetic voices. The findings provide insights into the state-of-the-art techniques for voice cloning and forgery detection.