Voice researchers say OpenAI’s 'Sky' sounds remarkably similar to Scarlett Johansson

OpenAI's recently discontinued voice model, codenamed "Sky," has sparked a humorous debate in the tech world. Researchers at Arizona State University, commissioned by NPR, compared Sky's voice to various actresses and found striking similarities to Scarlett Johansson's.

The Johansson Connection

The analysis, which reportedly used AI models to compare vocal characteristics, revealed Sky's voice shared a number of features with Johansson's, including:

• Vocal Tract Length: Both Sky and Johansson were found to have similar vocal tract lengths, which contribute to the overall sound quality and pitch of a voice.
• Articulation and Pronunciation: The researchers noted similar patterns in how Sky and Johansson articulate words and pronounce sounds.
• Overall Timbre: The analysis suggested Sky's voice, like Johansson's, possessed a certain richness and breathiness.

Intriguing, But Not Identical

While the research pointed to undeniable commonalities, it's important to note that Sky's voice wasn't an exact replica of Johansson's. The researchers themselves clarified that Sky likely represented a blend of various vocal characteristics, with Johansson being just one potential influence.

OpenAI Never Asked, Johansson Never Lent Her Voice

Adding another layer of intrigue, OpenAI reportedly never approached Scarlett Johansson about lending her voice for Sky. This raises questions about the specific data used to train Sky's voice model and the potential for using celebrity-like voices without consent.

A Fun Twist on AI Research

Despite the ethical questions, the Sky-Johansson connection injects a dose of humor into the often-serious world of AI research. It highlights the human element - our desire for familiar and pleasant-sounding voices - when designing artificial intelligence.

The Future of AI Voices

OpenAI's Sky project, though discontinued, represents a step forward in developing natural-sounding AI voices. As AI technology continues to evolve, it will be interesting to see how developers approach the creation of artificial voices, considering both technical capabilities and ethical implications.