Social media users have mixed opinions on the technology — some are excited to dub full TV shows, while others worry about the danger of fake news
A Taylor Swift deepfake video shows the American pop star speaking fluent Mandarin with ease.
Taylor Swift is a beloved musician in China, and the clip immediately went viral, racking up millions of views on social media and sparking discussion about the implications of AI dubbing technology.
The AI-generated videos were created using a tool from Chinese startup HeyGen. In the clips, Swift appears on a talk show to discuss her recent travels and musical inspirations — all in flawless Mandarin with perfect lip-syncing.
While many Chinese netizens expressed amazement at the realism of the technology, others raised concerns about its potential for misuse.
“I can’t imagine how this technology will be used in daily life in the future,” writes one Weibo user in a viral post. “The same person will speak different languages, depending on what region you’re in.”
“The scariest thing is using this to make fake news,” reads the top-rated comment under the post. “Because AI simulates the voice and mouth shape, it’s really easy for people to believe it.”
Others, meanwhile, were optimistic about the technology.
“Quick! Someone translate and dub all of ‘Game of Thrones,’” one user wrote.
Concern over the implications of convincing AI-powered deepfakes are not new.
In May, a high-profile case made headlines when a man was scammed out of 4.3 million RMB (around $570,000). Another major trend was the surge of interest that accompanied an AI impersonation of singer Stefanie Sun; Sun herself eventually came forward to comment on the technology, and she was not optimistic.
The Taylor Swift deepfake video, however, was even more advanced. Beyond a simple audio imitation, this was a blend of voice cloning, language translation, and visual dubbing of lip movements to create a singular, convincing clip.
In China, laws governing deepfakes are already in place. Regulations from the start of 2023 require AI-altered content to be labeled and tagged to avoid confusion. Additionally, creators are required to obtain consent from individuals being depicted, and businesses using deepfake technology are required to collect real-world identities from their users.
Nonetheless, policing and enforcing these rules remains a challenge.
Images via HeyGen
This post was last modified on November 10, 2023 9:49 pm