OpenAI’s latest chatbot system launch featured a natural-sounding voice named Sky, reminiscent of the AI character from the sci-fi film “Her”, voiced by Scarlett Johansson. This resemblance has sparked controversy, especially after it was revealed that OpenAI had approached Johansson to use her voice for GPT-4o, and she had declined.
Researchers have added fuel to the fire by finding that Sky’s voice is uncanny, similar to Johansson’s. National Public Radio (NPR) tasked researchers with comparing Sky’s voice to samples from over 600 professional actors. The results were striking: Sky’s voice was more similar to Johansson’s than to 98 percent of the other actors analyzed. OpenAI maintains that no intellectual property rights were violated, claiming an unidentified voice actor was used instead.
The researchers noted subtle differences: Sky’s voice is higher in pitch, breathier, and more expressive than Johansson’s usual tone. Occasionally, Sky’s voice matched more closely with samples from actors Anne Hathaway and Keri Russell. However, the scientists also measured specific vocal features, concluding that Sky and Johansson’s voices had identical measurements, which are influenced by physical aspects such as the throat, mouth, and nasal passages.
Despite the similarities, Visar Berisha of Arizona State University, who led the analysis, clarified that “the two voices are similar but likely not identical.” OpenAI CEO Sam Altman acknowledged the controversy at a recent conference, asserting, “It’s not her voice. It’s not supposed to be,” and apologized for the confusion. However, Altman’s single-word X post, “her,” during the GPT-4o event, deepened the association with Johansson’s role in Her. OpenAI CTO Mira Murati claimed unfamiliarity with Johansson’s voice until comparisons were made.
Johansson herself has reportedly been outraged by the similarity, which is particularly significant given her careful control over her voice’s commercial use. Her co-founder at The Outset, a skincare brand, revealed that Johansson had previously rejected voice simulation for customer outreach, underlining the personal and professional impact of the Sky voice resemblance.
In response to the backlash, OpenAI has pulled Sky, but the controversy remains unresolved. The incident underscores the broader implications of rapidly advancing voice cloning technology. While it holds promising applications, such as preserving the voices of those losing their speaking ability due to illness, it also poses significant risks. For instance, bad actors use voice cloning for phishing scams, manipulating users into divulging personal information.
The potential for misuse has led the SAG-AFTRA actor’s union to expand its standard contracts, emphasizing that only human actors can be credited as “voice actors” in animated TV shows and games. This move aims to protect intellectual property rights against the encroachment of AI-generated voices.
As voice cloning technology evolves, the balance between innovation and ethical considerations becomes increasingly crucial. OpenAI’s Sky controversy highlights the need for clear regulations and respectful use of such transformative technologies.
