Archival media remaster and artificial intelligence
Discover how artificial intelligence was used in the production to enhance storytelling and bring historical archives to life.
All remastered media are always shown alongside—or after—their original versions. For example, each 3D-remastered photo is presented next to the original photograph.
Storyboarding
Concept art and previsualization of the film
AI-assisted tools were used during the storyboarding phase to generate schematics, visual references, and pre-visualizations. These tools enabled us to explore different narrative structures, test camera movements, and design spatial environments in advance. This process helped the team plan shooting sessions efficiently, prepare each shot with precision, and even begin editing early drafts of the project.
Model used:
ChatGPT 4o for Prompt Generation
Luma Dream Machine
Stereoscopic conversion
3D conversion transforms 2D media into stereoscopic 3D by generating a separate image for each eye. In the past, this process relied heavily on time-consuming techniques such as manual rotoscoping and hand-painting.
Thanks to AI, we can now automate parts of this workflow, using neural networks to cut out elements and fill in missing parts. Artists remain essential to adjust depth and ensure the 3D result feels natural and realistic.
Using AI-generated depth maps, we brought select archival photographs into three-dimensional space. These maps assign spatial depth to each part of an image, allowing us to create immersive visuals from original 2D photos.
This technology was used on the following media:
Four final pictures taken by Richard Taylor on Omaha Beach
The seven photos in the interactive photo album
The two photographs of Richard Taylor holding his camera
Model used:
Adobe Photoshop Neural Filters
ImmersityAI 4.0 Neural Rendering
Nuke Cattery
Apple Vision Pro native stereoscopic conversion tool
Cleanup
Archival content often contains scracthes and film noise that can hinder visual clarity. We used AI-based denoising to remove these imperfections, making the images more readable for modern audiences—without compromising their historical integrity.
This technology was used on the following media:
The introductory archive footage from NARA
Richard Taylor’s footage visible in the viewfinder
Model used:
Topaz DeNoise AI
Neat Video
3D retexturing
To reconstruct realistic 3D scenes from archival material, 3D artists first modeled the underlying geometry and mapped textures from historical photographs to create an accurate base reconstruction.
When the resolution of the original images was too low, we used AI tools to generate higher-resolution textures based on visual references.
These AI-generated textures provided additional detail, which we then manually refined—when relevant, appropriate, and historically accurate—to blend seamlessly with the real scenes. The result: lifelike recreations that remain faithful to the original artists’ intentions:
This technology was used on the following media:
3D scene of soldiers leaving the landing craft
3D scene when landing on the Omaha Beach
3D scene with soldiers running on Omaha Beach
Model used:
MagnificAI
ChatGPT 4o image generator