Deepfake Voice System for Spoofing

Abstract

This paper presents a deep fake voice spoofing system. To develop this system, the concept of a voice modulation system for altering and replicating human voices are explored. Tacotron2, Glow-TTS, and XTTS-V2 are studied, and they demonstrate the significant advancements in voice replication technology. The methodology involves a comprehensive process of audio data collection, and model implementation, followed by integration into a user-friendly website. The evaluation phase employs user acceptance testing with the Mean Opinion Score (MOS) as the primary metric, where participants rate the quality and resemblance of the synthesized voice compared to the original. The results highlight the voice modulator's effectiveness in achieving high-fidelity voice replication, with positive user feedback.