Variety reports As Deep as the Grave used generative AI for Val Kilmer's performance

• Variety’s exclusive report, surfaced in the primary thread, says As Deep as the Grave used generative AI to create Val Kilmer’s performance even though he never filmed scenes for the movie.
• According to the report thread, the filmmakers used footage and images supplied by Kilmer’s family, and the story says his estate and family backed the release.
• The first-look post frames this as a substantial on-screen role, not a quick cameo, making it an early high-profile case of consent-based digital resurrection in a feature film.
• One concrete production detail in Variety’s writeup is that the AI performance preserved Kilmer’s real speech limitations from throat cancer rather than replacing them with a cleaned-up synthetic version.

The core filmmaking move here is specific: As Deep as the Grave did not use archival footage as-is. Per Variety’s report, director Coerte Voorhees and his team generated Kilmer’s likeness and voice for the character Father Fintan after production delays, drawing on materials provided by Mercedes and Jack Kilmer. The article says scenes that had been at risk of being cut for budget reasons were brought back through that process.

That makes this less a novelty cameo than a post-production performance build. The first-look post points to a finished character presentation in costume and scene context, which matters for filmmakers because the creative choice extends beyond face replacement into performance continuity, character blocking, and audience acceptance.

For directors and VFX teams, the most usable takeaway is the consent structure. In a reply in the discussion, one creator argues that family approval can make this kind of work feel restorative rather than extractive, and Variety’s article makes clear that family cooperation was central to the production.

The harder creative question is tone. The film reportedly kept Kilmer’s illness-shaped voice in the character instead of generating a frictionless approximation, which suggests a more documentary-minded use of generative tools than a pure “de-aging” fantasy. That choice will likely shape how audiences judge similar AI performances going forward.