OpenAI has developed a powerful new AI tool called Voice Engine that can generate highly realistic voice clones from just 15 seconds of recorded audio. However, the company is hesitant to release it widely due to concerns over potential misuse for spreading misinformation, especially during election years.
While an initial version powers the text-to-speech feature in ChatGPT, the full capabilities of Voice Engine have not been publicly unveiled. OpenAI aims to initiate a dialogue around the responsible deployment of such synthetic voice technology and how society can adapt.
The AI lab shared examples of Voice Engine’s applications from select partners, including educational voiceovers, accented voice translations, and even restoring the voice of a woman who lost her ability to speak due to a brain tumor.
Despite the benefits, OpenAI believes it’s crucial to build “societal resilience” against deceptive AI content before unleashing Voice Engine on a larger scale. The company suggests phasing out voice authentication for sensitive data and exploring policies to protect individuals’ voices.
Educating the public on AI’s capabilities and limitations is also vital, as is watermarking generated audio to trace its origin. While OpenAI requires consent from original speakers, some competitors already offer public voice cloning with election safeguards.
As synthetic voice tech advances, balancing its benefits with mitigating risks like misinformation will be paramount for responsible deployment.
