The process of using digital effects to change the age of actors is becoming increasingly popular in Hollywood many criticisms), but to everyone unusually young Robert DeNiro Irishman, requires hundreds of hours of hard work behind the scenes from VFX artists and engineers. etc neural network while becoming more widespread and accessible, heavy hitters in the entertainment industry are now able to use the technology to streamline processes to simultaneously create more photorealistic results. Few entertainment companies are as popular as Disney right now, so it’s no surprise to hear that the House of Mouse recently announced it’s using cutting-edge artificial intelligence to “re-age” its actors.
yesterday, Disney Research Studios unveiled Face Re-Aging Network (or FRAN), the latest advancement in VFX work that uses neural networks to create “the first practical, fully automated, production-ready method for facial re-aging in video footage.” according to Disney’s own numbers, in 2022 at least 12 movies and TV series have used aging technology, a number that has grown steadily since its debut a few years ago. “Photo-realistic digital aging of faces in video is becoming increasingly popular in entertainment and advertising. But mainstream 2D drawing often requires frame-by-frame manual work that can take even skilled artists days,” the team summarizes. their research paper. FRAN’s solution to the problem is based on a multi-stage neural network process, which improves the realism of the effect while reducing labor time and cost.
Watch Disney video review below:
etc Gizmodo and Ars Technica As explained earlier today, FRAN uses several steps to create its aging themes. First, a program called team was used StyleGAN2 Randomly generate thousands of synthetic aged faces between the ages of 18 and 85. After this database was created, machine learning tools aged, aged, and then de-aged these artificial portraits. different neural networks, FRAN. From there, FRAN can apply what it’s learned to videos of real people, adding and subtracting years regardless of angle, position, movement, or lighting. Artists can then step in and manually edit frames and transitions as needed.
The results are still not perfect, but they are still as believable as in the movies. Thief One or Ant-Man and The Wasp. If nothing else, neural networks like FRAN will become technically smarter and the end product will improve over time.