• January 27, 2023

Artificial Intelligence

Inflection Point Artificial Intelligence (AI) is the ability of machines to perform tasks that would normally require human intelligence. The rise in computer power, along with digital data, is what makes …

New Samsung Leak Reveals Galaxy S23 Price Shock

Samsung’s Galaxy S23 range delivers the most anticipated Android smartphones of the year, but international buyers are in for a shock when they see the prices. A new leak from Roland …

New Apple Leak Reveals Cutting-Edge iPhone 15 Feature

iPhone 15 leaks have revealed many of the phone’s biggest upgrades. Now a new report claims Apple will equip models with Wi-Fi 6E, described as “the most disruptive boon for Wi-Fi …

FRAN might just become Disney video editors’ new best friend.

New research by Disney Research Studios, which focuses on technology innovations for filmmaking, shows off a new face aging network, named FRAN, that simplifies the aging process of actors in video.

Thousands of fictional faces were generated by Disney researchers based in Zürich and input into a database, training the U-Net architecture on how to age and de-age them. FRAN can predict which parts of the face will change with age and can automatically generate or remove wrinkles, for example.

By layering these details over the original face, the resulting shot will also retain more of the actor’s real appearance. Additionally, artists will be able to fine tune the effect of aging with more localized control and creative freedom thanks to the AI.

Disney did note, however, 3D re-aging solutions may be preferable for more elaborate levels of control, as FRAN is based on a 2D re-aging workflow.

Advertisement

Photorealistic digital re-aging is quite common, but the process has been fairly tedious to date. In the past, much of this work was done manually by artists, who could take weeks to complete it. It is sometimes necessary to reshoot sections of a film in order to be able to re-age it, which can come at hefty costs to Hollywood studios.

For a while now, AI has been used to re-age still and video images. In moving video, the subject is often exposed to different variables every time he or she appears, making the re-aging of still images using AI difficult. Video re-aging AI solutions often feature ghosting and other temporal artifacts, resulting in poor video quality and a lack of photorealism.

Using this AI on a mass scale could have benefits beyond filmmaking. In 2019, four North American universities revealed a similar technology. One co-author of the research told Fast Company the AI could be used in missing child cases.

Disney noted in its research that re-aging from very young ages is challenging using U-Net, due to how many image changes are required, recommending LATS or DLFS for that use case.

Researchers also noted that the greying of scalp hair was not captured in FRAN’s current training data and is not reflected in outputs.

It is unclear when the technology, if ever, will be released to the public.

Advertisement

Leave a Reply

Your email address will not be published.