“The company is exploring ways of watermarking synthetic voices or adding controls that prevent people from using the technology with the voices of politicians or other prominent figures.”
– Cade Metz
The latest in a string of generative AI models from OpenAI can replicate your voice based on 15 seconds of training text you read to the system. The company is holding back a general release of these newest capabilities so potential harms can be better understood and mitigated.
The New York Times reports that the model, currently called Voice Engine, can say any text it’s given, including translation into other languages, after being trained on a short passage read by the speaker.
OpenAI Unveils A.I. Technology That Recreates Human Voices | THE NEW YORK TIMES | March 29, 2024 | by Cade Metz