OpenAI has recently introduced a new technology called Voice Engine, which has the ability to generate natural speech that closely resembles the original speaker in a fifteen-second audio clip. The tool is capable of replicating voices in four different languages including English, Spanish, French, and Chinese. While the Voice Engine was developed by OpenAI back in 2022, the company has been careful in its release and aims to spark a discussion on responsible deployment of synthetic voices.
The Voice Engine has the potential to be utilized in a variety of ways, including providing reading assistance to non-readers and children by generating pre-scripted voice over content with natural-sounding voices. Another application is helping patients recover their voices in cases of sudden speech loss or degenerative speech conditions. Brown University has been testing Voice Engine to aid patients with oncologic or neurologic issues that impact their speech function.
Partners that are testing Voice Engine are required to adhere to OpenAI’s usage policies, which prevent the impersonation of individuals or organizations without consent. The original speaker must provide explicit and informed consent, and developers are not allowed to enable users to create their own voices. Additionally, partners must disclose that the voices generated are AI-driven, and OpenAI is implementing watermarking to track the origin of any audio produced by Voice Engine and to monitor its usage proactively. Despite its potential benefits, Voice Engine also poses risks such as fraudulent extortion scams, false election and marketing campaigns, and potential misuse of voice artists’ work.
OpenAI has suggested several safety measures for voice technologies, including phasing out voice-based authentication for sensitive information, protecting individuals’ voices in AI, educating the public about AI capabilities and limitations, and developing techniques to track the origin of audiovisual content. The company has decided to delay a formal public release of Voice Engine due to safety concerns, citing the election year as a factor. It has also filed a trademark application, hinting at possible competition with Amazon’s Alexa in the market. Voice cloning technology appears to be a lasting trend, and OpenAI aims to position itself as a key player in this field.
In related news, the Federal Communications Commission (FCC) announced that calls made using voices generated with AI technology will be classified as “artificial” under the Telephone Consumer Protection Act (TCPA). This decision makes it illegal for robocalls utilizing voice cloning technology to target consumers, offering protection against potential misuse of AI-generated voices in phone communication. Overall, the development of Voice Engine and the regulatory updates reflect the continued advancement and evolving landscape of AI-driven technologies in communication and security sectors.
