Disney Research has a new visual effects tool that can make actors look older or younger for movies without needing to collect longitudinal data using photographs.
A paper recently presented at ACM SIGGRAPH Asia (2022) introduced FRAN, a "Face Re-Aging Network" that can:
automatically re-age face images in video without identity loss
Intended for visual effects, FRAN comes from a team of researchers from the Zurich-based Disney Research Studios, and ETH Zurich. Setting the scene the paper states:
The use of digitally aged or de-aged human characters in movie productions and advertising has increased dramatically over the last few years. Whether for making Robert De Niro younger in The Irishman [2020], or David Beckham older in an ad campaign against malaria [2020], digital re-aging is quickly becoming an essential tool in every visual effects studio.
This video presents the research, opening with video frames of an male aged 35, re-aged to 65 (top row) and to 18 (bottom row):
As an AI-based, fully-automatic tool, FRAN is intended to replace the predominant 2D painting workflow which often requires frame-by-frame manual work that can take days to accomplish, even by skilled artists.The research paper states:
Although research on facial image reaging has attempted to automate and solve this problem, current techniques are of little practical use as they typically suffer from facial identity loss, poor resolution, and unstable results across subsequent video frames.
The approach used for FRAN that overcomes these problems dispensed with collecting data from real human subjects. Instead the neural network was trained with a large database containing pairs of randomly generated synthetic faces at varying ages.
FRAN also provides artists with localized control and creative freedom to direct and fine-tune the re-aging effect, such as adding or erasing wrinkles and jowls, something the paper suggests is important in real production pipelines and often overlooked in related research work.
The results seem impressively credible but it isn’t clear if Disney intends to make this technology available to the public.
Google is making a new differential privacy library available as open source. PipelineDP4J is a Java-based library that can be used to analyse data sets while preserving privacy.
TestSprite has announced an early access beta program for its end-to-end QA tool, along with $1.5 million pre-seed funding aimed at accelerating product development, expanding the team, and scaling op [ ... ]
