New Short ‘Reflection Of A Reflection Of A Reflection’ Was Produced Using Images Generated By Stable Diffusion

Germany-based Italian filmmaker Fabian Mosele is a filmmaker who experiments heavily with the use of machine learning in his work. The images used in his latest piece Reflection of a Reflection of a Reflection were generated using Stable Diffusion AI image-generating software.

In the film, Mosele presents a loosely narrative story inspired by the online blog post “I, We, Waluigi: a Post-Modern analysis of Waluigi.”

In his own words, Mosele explains, “My latest film is an ode to Waluigi, reflecting on how generative media can shape our individuality. It’s one of the first animated films assisted by image generative models for the visuals.”

Aesthetically, there are clear Mario universe influences to be found in many of the film’s characters, especially during a picnic scene near the end. While there are visual incongruities and a disconcerting quality that are the hallmarks of current AI-generated images, there is also a greater continuity from shot to shot than can be found in many films which heavily feature AI artwork.

That is largely due to the fact that Mosele put in a lot of human hours on this piece as well. In fact, production began in October 2022 and finished in May 2023. In a detailed making-of video that accompanies the short, Mosele explains how much human labor was required to put the film together:

Mosele’s film demonstrates both the limitations of using AI software in animation and the possibilities it affords artists who want to experiment with the technology.

For those who are interested in exploring further how Mosele crafted the film, he has made available the assets used to produce the film including the Stable Diffusion models and the parameters and prompts that created each scene.