In with the Old
A video was posted to YouTube today by Denis Shiryaev showing how he used Neural Network AI to upscale historic footage, in this case L’arrivée d’un train en gare de La Ciotat (1896). While similar AI techniques recently gained attention through Deepfakes, upscaling older footage has promising possibilities for film preservation—though it raises concerns about authenticity.
Personally, I love the idea of upscaling while still aiming for fidelity, but changing the framerate can resemble motion smoothing, often clashing with a filmmaker’s intent. It reminds me of the controversy in the 1980s when Turner Entertainment introduced film colorization.
Would the Lumière Brothers have opposed such manipulations of their film? Perhaps not—but that doesn’t mean we should do anything more than respectfully restore classic cinema. The documentary Lumière! (2016) was a much more faithful approach, offering beautiful restorations. As for creative experimentation, it demonstrates that cinema can be changed in any way we can imagine.
Links
Films
- L’arrivée d’un train à La Ciotat (1896)
[IMDb // Letterboxd] - Lumière! (2016)
[IMDb // Letterboxd]
References
- Artificial Neural Network
- Deepfake
- Film Colorization
- Motion Interpolation
- Turner Entertainment
- Video Scaler