Augmenting a person’s voice has been around for decades. I remember watching Unsolved Mysteries as a child. Sometimes the show masked a persons’ voice to conceal an identity. A baritone voice on a man who is an alto obscured the average viewer from knowing him. Voice cloning is not the same, but it can mask an individual in a comparable way. From another angle it looks like someone is pretending to be another person.
All things should be ethical
Whether or not the donor voice has prior consent means everything. Remain in line with the Association for Computing Machinery Code of Ethics 3.1 is to “Ensure that the public good is the central concern during all professional computing work.”
The replication of vibrations for each human and their larynx should require consent.
Generative Adversarial Networks (GANs), created by Ian Goodfellow, is one tool to harness to make quality audio. Other methods are with Python libraries such as TensorFlow, PyTorch, and Librosa. The No Starch Press Book called Make Python Talk helps readers learn methods to build apps with voice control and speech recognition.
The first GitHub Repository allows the user to clone a voice in five (5) seconds and has a graphical user interface to make it easier. The second one allows text-to-speech synthesis and can be for real-time voice cloning.
Knowing how to use Python Libraries will come in handy for both repositories.
To supplement these repos, look to these Medium User articles.
How To Clone Voices with Python + PyTorch
Adversarial Audio Synthesis
A proprietary application with free and paid options are available. The purpose is not to mimic someone else, but to transform text into your voice. Make sure to acquire consent, if you use someone else's likeness.