OpenAI Unveils AI Voice Generator, Raising Concerns and Potential for Misuse
OpenAI has introduced a groundbreaking artificial intelligence tool named Voice Engine, capable of mimicking human voices with remarkable precision. While the tool holds promise for accessibility services, it also raises concerns about misinformation and potential abuse.
The Voice Engine utilizes a 15-second sample of human speech to generate lifelike replicas of voices. Users can input text, and the tool reads it aloud in the AI-generated voice, as demonstrated in early tests shared by OpenAI on Friday.
Although various AI-generated voice services already exist, OpenAI’s track record, notably with ChatGPT, suggests its potential for widespread adoption. The applications for such a text-to-voice tool range from translation services to aiding individuals with speech impairments. However, skeptics fear it could exacerbate disinformation and facilitate scams.
Currently, Voice Engine is restricted to a select group of trusted partners, including educational and health technology firms. Testers must obtain explicit consent before replicating voices and must clearly disclose that the voice is AI-generated. OpenAI acknowledges the risks associated with synthetic voice technology, particularly in sensitive contexts like elections.
While Voice Engine is not yet available to the public, OpenAI plans major changes before any broad deployment. These may include phasing out voice-based authentication for sensitive accounts and implementing voice authentication to verify the original speaker’s consent.
Voice Engine boasts multilingual capabilities, allowing it to generate voices in various languages while preserving the tone and accent of the original speaker. OpenAI’s blog post showcases samples of AI-generated audio reading the same passage in multiple languages.
The announcement of Voice Engine precedes the anticipated public release of Sora, an AI-generated video tool teased by OpenAI. Sora creates realistic 60-second videos from text instructions, complementing ChatGPT’s ability to generate images from text prompts.
Additionally, OpenAI has made ChatGPT available to the public without requiring sign-up, albeit with limitations on certain features. Users can opt-out of data usage for model improvement, but without an account, they cannot access voice conversations or save chat history.
