Disney Researchers Created An Artificial Intelligence Tool (AI), That Instantly Makes An Actor Look Younger or Older in A Scene

Performance results

FRAN can be applied to video frames and produces reliable re-aging results. It is flexible enough to adapt to changes in head position, lighting and depth of field.
FRAN consistently and convincingly re-ages images supplied while keeping the identity of the target. The timeline shows that the aging process is both continuous and linear.
FRAN is more accurate than HRFAE in that it preserves the individual’s skin details and input identification.
FRAN’s processing of real-world photos performs nearly as well as those simulated.
FRAN is unlikely to replace many industry jobs in the near future, as manual VFX and actual prosthetic makeup applications are not subject to these limitations. There are some limitations to FRAN, and these are not new studies. Disney discovered that FRAN was not ideal for making drastic changes such as re-aging at very early ages. Also, the greying of the scalp hair wasn’t taken into account when an actor is aging up because it wasn’t part of the data used to train FRAN.

To summarize

Visual effects can be made simpler by making them easier. This includes reducing the workload for already underpaid and overworked artists and making it more affordable to filmmakers with smaller budgets. The ability to automate labor is a business motivation even for large studios. Businesses like Disney invest in research to improve visual effects quality. In recent years, some of this research have focused on AI and how it might streamline the process.

There are many reasons Disney would want to make such a device. This could save visual effects artists time and reduce the amount of work required to complete their projects. It could help lower-budget movies artificially age their actors while reducing production costs. When applied to still images, the results of neural networks and ML are convincing. However, they are not photorealistic when applied in moving video. There are temporal artifacts that appear and disappear from frame to frame, and sometimes the appearance of a person becomes unrecognizable as the video is altered. These fictitious individuals were aged using machine learning aging methods. The data were used to train a new neural network called FRAN (face-re-aging network).