OpenAI, creators of the renowned ChatGPT, has unveiled a remarkable voice-cloning tool called Voice Engine. This potent system can mimic a person’s voice, including emotion and natural speech patterns, using just a 15-second audio sample.
Due to the immense potential for misuse, OpenAI is taking a measured approach in releasing this technology. Voice Engine could easily create convincing deepfakes for malicious purposes.
OpenAI is currently testing the tool with select partners and sees potential benefits. Voice Engine could assist with reading for children, translate content with an individual’s own voice, and even give a voice to people who have lost theirs.
To mitigate risks, OpenAI insists partners clearly label AI-generated audio and has built in a watermarking system. However, a widespread public release isn’t guaranteed.
OpenAI acknowledges the need for a broad conversation about responsible use of synthetic voices. The company’s decisions on Voice Engine’s distribution will be heavily influenced by how society reacts to this powerful new form of generative AI.