ReVoice: A Neural Network based Voice Cloning System

Voice cloning, a transformative technology, aims to synthesize speech in a target voice by using provided text and a few audio samples of the target speaker. In recent years, voice cloning has witnessed extraordinary advancements, driven by the application of deep learning techniques. This technology has opened up innovative opportunities for human-computer interaction, personalization, and content creation. The purpose of this paper is to introduce a personalized voice cloning system that utilizes deep learning models for Text-to-speech (TTS) synthesis and audio generation. Subsequently, users can synthesize speech by inputting text and providing an audio sample. By leveraging deep learning techniques, this system captures the distinctive vocal characteristics of the user and employs them to create natural-sounding speech output. There are numerous emerging use cases where voice cloning can be applied, including actors using it to dub their movies in different languages, people who have lost their voices using this technology to communicate, and advertisers creating multiple versions of ad reads from one script to avoid sounding repetitive. Additionally, this technology has opened the door for numerous companies to offer new services and products. It can be used to create voices for chatbots, audiobooks, video games, text readers, and more.