Showrunner wants to use generative AI to recreate lost footage from an Orson Welles classic

Showrunner, an AI startup, plans to use generative AI to recreate the lost 43 minutes of Orson Welles' 1942 film, The Magnificent Ambersons. This project aims to restore Welles' original 131-minute vision, which was cut down to 88 minutes by RKO without his consent, and is presented as an academic endeavor rather than for profit.

QUICK TAKEAWAYS

• Showrunner aims to restore lost footage of Orson Welles' The Magnificent Ambersons using generative AI.
• The project seeks to recreate the original 131-minute cut, disowned by Welles after studio edits.
• AI will generate approximations of missing scenes, using keyframes and manipulating live actors' faces.
• Collaborators include AI VFX artist Tom Clive and filmmaker Brian Rose, who previously attempted a similar restoration.
• The company emphasizes the project is non-monetized, aiming to legitimize AI in entertainment and fulfill a historical cinematic quest.

KEY POINTS

• Orson Welles' The Magnificent Ambersons was originally 131 minutes but was cut to 88 minutes by RKO, leading Welles to disown the film.
• Showrunner is developing a new generative AI model, including its FILM-1 model, to approximate the lost footage based on Welles' notes, shooting scripts, and archival photos.
• The AI approach involves generating keyframes for missing scenes, using set photos for spatial settings, and manipulating live actors' faces to resemble the original cast.
• The team includes AI VFX artist Tom Clive, known for face-swapping work on films like Alien: Romulus, and filmmaker Brian Rose, who has prior experience attempting an Ambersons reconstruction.
• Despite previous instances of creating unauthorized AI-generated content (e.g., South Park episodes), Showrunner states this project is not for commercial gain but to see Welles' original vision exist.

PRACTICAL INSIGHTS

• AI Models/Tools: Showrunner's proprietary generative AI model and its FILM-1 model are central to the recreation.
• Methods: Combines AI-generated approximations, use of original set photos, and AI-manipulated live actors for character representation.
• Talent: Leverages specialized AI VFX artists and filmmakers experienced in film reconstruction.
• Monetization Strategy: The project is explicitly non-monetized, framed as an academic and restorative effort.

PRACTICAL APPLICATION

This initiative demonstrates a novel application of generative AI in film restoration, potentially enabling the recovery of lost or incomplete cinematic works. It highlights how AI can be used to bridge historical gaps in media, offering new avenues for cultural preservation. However, it also underscores ongoing challenges related to intellectual property rights, especially when using AI to recreate content based on copyrighted material, even with a stated non-commercial intent. The success and reception of this project could set a precedent for future AI involvement in archiving and re-envisioning classic films.