"If You Hear My Voice, It Mimics Exactly"…OpenAI Unveils New Technology

OpenAI has developed and released an artificial intelligence (AI) tool that learns and mimics human voices. While there is growing anticipation that the rapidly evolving AI technology will bring innovation to humanity, concerns about voice manipulation are also emerging.

Sam Altman, CEO of OpenAI

On the 29th (local time), OpenAI revealed preliminary experimental results of its developed 'Voice Engine' on its blog. The company stated that with just a 15-second voice sample, the Voice Engine can generate a voice similar to the original speaker's voice. OpenAI first developed the Voice Engine at the end of 2022.

They explained that it was used to enhance voice functions in ChatGPT’s speech recognition and reading features, as well as in the text-to-speech conversion API (application programming interface). The actual human voice samples and the voices generated by the Voice Engine released by OpenAI on this day were almost indistinguishable.

OpenAI described that this tool can be positively utilized in various fields. Representative examples include voice narration content for children's education, real-time personalized response generation, and services that translate content such as videos and podcasts into multiple languages for global distribution. Additionally, OpenAI reported cases where it has been applied in therapeutic applications for patients with diseases affecting language functions and communication devices for people with disabilities.

Bloomberg News reported, "OpenAI has opened a new horizon in AI technology by unveiling the ability to mimic human voices," adding, "This also raises concerns about the risks of deepfakes (AI-generated videos, images, or voice manipulations)." OpenAI stated, "Due to the potential misuse of synthetic voice functions, we are approaching broader release cautiously," and "For now, we will only provide a preview of this technology without a wide release."

OpenAI added, "Ultimately, whether we widely distribute this technology or not, it is important for people around the world to understand where this technology is headed," and "We look forward to continuing the conversation about the challenges and opportunities of synthetic voices with policymakers, researchers, and developers."