Disney's Artificial Intelligence (AI) Tool
Reading Time: 3 minutes

Disney researchers have developed an Artificial intelligence tool that makes actors look younger or older in a scene. 

Over the years the entertainment and advertisement industry has been using photorealistic digital re-aging of faces in video. This process is relatively time-consuming and tedious. 

Researchers at Disney have been trying to unlock the potential of artificial intelligence in photo-realistically changing footage. Their revolutionary creation is an age-shifting device that can make an actor seem convincingly older or younger without weeks of costly, specialized CGI labor.

Disney’s research and development department has developed an AI system to make actors appear their age in various scenes. To get the best results, artists still edit individual frames by hand, but the AI tool can do a lot of the work for them. One way this could be significant is that an individual frame containing the effects of aging might require no more than five seconds of processing time…

Synthesizing high-quality, longitudinal aging data is a method FRAN uses to face re-age visual effects. This technique involves devising a way to avoid the seemingly impossible challenge of collecting annotated pictures of people (people with diverse ages, races, and genders). This technique produces several input-output collections that show the same person at two different ages with consistent facial expressions, stances, lighting, and backdrops.

How does Disney’s Artificial Intelligence (AI) Tool work?

Researchers carefully adapted the U-Net design to improve translation quality and re-aging control, given the image-to-image translation problem. Multiplied by L1, perceptual, and adversarial losses during training, FRAN is built from paired synthetic data. With the input image converted into a single-channel age map, FRAN then predicts per-pixel RGB changes (offsets) that layer on top of the original picture. In most cases, FRAN takes input from video frames and outputs good temporal consistency. These features work together to create FRAN.

Disney’s Artificial Intelligence (AI) Tool Findings

  1. Applying FRAN on a video frame can create a reliable re-aging effect, and it’s able to automatically adapt fluidly when lighting conditions change and we gain more insight into the subject.
  2. With FRAN, images are re-aged believably and convincingly. When you look at the timeline, the process is as continuous and linear as possible.
  3. In short, FRAN achieves clear facial recognition, whereas HRFAE does not.
  4. FRAN has found that real-world photographs have the same working quality as those simulated ones. This means less work for you and more time to focus on other aspects of your business.

The introduction of FRAN will not displace industry jobs for the time being while the manual VFX work. Also, the actual prosthetic makeup application has no such limitations. Though there are a few drawbacks as FRAN isn’t perfect for drastic changes like going from being extremely young to suddenly being old, or aging someone to an age when their hair turns grey, according to Disney. 

Visual effects have become much simpler over time, which has a number of benefits. Not only does this save time for the already overworked and underpaid artists, it helps filmmakers with small budgets produce higher-quality content. Another benefit is that some AI companies are working to improve the technology even further. For example, Disney funds study exploring this area and in recent years some of these studies have been about how AI can streamline the process.

Disney will certainly benefit from its creation as it will help them in many of their projects. It will reduce the cost of production considerably. Though there are shortcomings when it comes to the use of AI and ML on moving images. Temporal artifacts occur and it can look like the subject is changing ages from frame to frame. As a result, hundreds of people have been aged using pre-existing ML training, which was then used to train a new neural network.

Related Articles:
Oneka Wave-Powered Buoys Reduce the Ecological Cost of Desalination Drastically
3D Microscope Video Tracks Virus Whizzing Around in Real Time
Dancing Rats Shed Light on the Ability of Animals to Move to a Beat