AI Rewrites History: Reconstructing Welles' Lost Vision in 80 Years

A pioneering project is set to use artificial intelligence to reconstruct lost footage from Orson Welles’ 1941 film The Magnificent Ambersons, marking a significant intersection of AI and filmmaking. The initiative is led by Showrunner, a tech startup backed by Amazon, which recently announced the development of an AI model capable of generating long-form, live-action films. The technology is intended to allow users to create custom episodes of TV shows with minimal input, but in this case, it will be used to recover 43 minutes of the original film that were reportedly destroyed by studio executives. The original film was 131 minutes long, but after editing by RKO Pictures, it was reduced to 87 minutes, a decision that sparked controversy and left many to speculate about the artistic value of the missing scenes [1].

The project represents a broader shift in the entertainment industry as AI begins to encroach upon creative domains traditionally reserved for human talent. Showrunner’s CEO, Edward Saatchi, envisions the platform evolving into what he calls the “Netflix of AI,” where users can interact with intellectual property and even create their own versions of existing content. While the current AI tools can only sustain a story for a short episode, Saatchi sees the technology moving rapidly toward the ability to generate full-length films. This development has raised concerns about the use of copyrighted materials in training AI models, as well as the potential for such tools to displace human creators in the future [1].

The reconstruction of The Magnificent Ambersons is not intended as a commercial venture. Showrunner does not hold the rights to the film, which are owned by Warner Bros. Discovery and Concord. The goal, according to Saatchi, is to bring the lost footage into the world and allow it to be seen after eight decades of speculation about its artistic merit. The company hopes to spark a broader conversation about the original vision of Welles, who was known for his creative control and innovative techniques in early Hollywood [1].

The project will employ a blend of traditional film techniques and cutting-edge AI technology. Some sequences will be shot with live actors, and AI tools will be used to transfer facial expressions and poses to preserve the likenesses of the original performers. Extensive archival materials, including set photos and 3D models of the original sets, will be used to reconstruct the scenes accurately. Brian Rose, a filmmaker who has spent five years working on the project, has already re-created 30,000 missing frames. His work includes mapping out camera movements and framing decisions to match the original script and archival footage [1].

The effort also involves Tom Clive, a visual effects expert known for his work in face-swapping and de-aging technology. Clive’s experience will be crucial in ensuring the reconstructed scenes maintain the aesthetic and emotional impact of the original film. One particularly notable scene that is being restored is a four-minute-long, unbroken moving camera shot that showcased Welles’ advanced filmmaking techniques. The loss of most of this sequence has been described as a tragedy by Rose, who emphasizes its complexity and narrative significance [1].

The initiative underscores both the potential and the controversy surrounding AI in the arts. While it offers new ways to preserve and restore lost cultural works, it also raises ethical questions about the role of AI in creative fields. As the technology continues to evolve, the film industry faces the challenge of balancing innovation with the value of human artistry [1].

Source: [1] Orson Welles' Lost Movie Will Use AI to Reconstruct (https://www.hollywoodreporter.com/business/business-news/orson-welles-lost-movie-ai-1236361881/)