Create synthetic human voices with Haunted, an open-source behavioral synthesizer, and generate natural-sounding audio files from a small voice sample
Haunted is an open-source behavioral synthesizer that allows users to create synthetic human voices. It only requires a small sample of a person's voice to then generate natural, human-like speech reading out any text the user inputs.
The software is based on recent advances in AI techniques such as generative adversarial networks and autoencoders. It can capture the timbre, rhythm, and intonation of a voice from as little as 30 seconds of sample audio. Users simply record a short voice sample and then can type in any text for Haunted to read out loud in that same voice.
Some key features and capabilities of Haunted include:
Haunted provides an easy way for anyone to clone a voice and create synthesized speech. It has many creative applications and continues to be developed as an open-source project to push the boundaries of what’s possible with AI voice generation.