Animating Ancient Painting: Research and Experimentation for Le Clarisse
This project was born from the desire to explore the potential of generative animation using artificial intelligence algorithms applied to historical and artistic heritage.
In collaboration with the Polo Culturale Le Clarisse in Grosseto, and based on an idea by th Museum Director Mauro Papa—who supervised all phases—I created a series of short animations based on paintings from the Gianfranco Luzzetti Collection, viewable in the museum via QR codes installed next to the original works.
The work required an in-depth study phase: each animation stemmed from an iconographic and compositional analysis of the artwork, followed by the crafting of specific prompts, designed to stay as faithful as possible to the spirit, structure, and theological or narrative meaning of the painting.
The goal was not to “modernize” the images, but to activate what is already dynamic and narrative within the painting, while respecting the limits of the original material and the available technologies.
The aim was also to spark curiosity in a younger audience by communicating art and cultural spaces in a contemporary form—without distorting them.
Technical and Ethical Challenges
Working with historical works using AI generation tools raises complex questions. The platforms used (often opaque in their internal processes) have structural limitations that make it difficult to fully control the resulting image.
One clear obstacle was the inability to faithfully animate the body of the nude Baby Jesus due to automated filters, which—though intended to prevent inappropriate content—fail to distinguish sacred from improper contexts.
Other recurring issues included the inconsistent or distorted rendering of hands, instability in the facial features of secondary characters, and the appearance of “hallucinated” elements not present in the original painting.
A significant part of the work, therefore, involved developing compensation strategies to preserve visual coherence and iconographic fidelity within technical constraints.
An Ongoing Journey
This project does not aim to offer “definitive” animated versions of the works, but rather to open up a space for exploration—between preservation, interpretation, and technological experimentation.
The museum thus becomes a place where digital tools do not replace but complement direct observation, suggesting new ways of interpreting artworks without intruding on their original meaning.
In these pages, you’ll find the animations, images of the original artworks, and a detailed section on the prompts used. I’ve also included a separate section dedicated to errors and discarded results—because these too are part of the research process.
4o
Animations with AI
This section collects animations generated through the use of artificial intelligence, starting from selected works from the Luzzetti Collection.
Each animation is born from a work of visual analysis, iconographic study and writing of accurate prompts, with the aim of respecting the style, composition and original meaning of the painting.
The animations are not intended to replace the work, but to offer a dynamic reading opportunity that stimulates curiosity and opens new interpretative paths, especially for younger audiences.
Errors and hallucinations
During the process, the use of generative platforms produced numerous unexpected results: distortions, incorrect anatomies, added or inconsistent elements with respect to the original work.
This section collects some of these “visual scraps”, not for irony, but to document the current limits of technology and make the process transparent.
The errors help to better understand the relationship between historical image and artificial intelligence, and are an integral part of this research.
Here you can find some of the animations with errors and hallucinations.
Animating Ancient Painting: Research and Experimentation for Le Clarisse
This project was born out of a desire to explore the potential of generative animation, using artificial intelligence algorithms, applied to the field of historical and artistic heritage. In collaboration with the Polo Culturale Le Clarisse in Grosseto, and based on an idea by Mauro Papa, who supervised every phase of the process, I created a series of short animations based on paintings from the Gianfranco Luzzetti Collection, now viewable in the museum via QR codes placed next to the original artworks.
The project required an in-depth research phase: each animation was developed from a visual and iconographic analysis of the painting, followed by the careful writing of specific prompts designed to remain as faithful as possible to the original spirit, structure, and theological or narrative meaning of the work.
The goal was not to “modernize” the images, but rather to activate what is already narrative and dynamic within the painting itself, working within the boundaries set by both the original medium and the current technological tools.
Another objective was to spark curiosity in a younger audience, offering a more contemporary form of engagement with art and cultural spaces without distorting their essence.
Technical and Ethical Challenges
Working with historical artworks through AI-based generative tools raises several complex issues. The platforms available (often opaque in their internal logic) present structural limitations that make full control over the image difficult. A significant obstacle was the inability to faithfully animate the naked body of the infant Jesus, due to automated filters that, while intended to block inappropriate content, fail to distinguish between sacred and profane contexts.
Other recurring issues included the inaccurate rendering of hands, facial instability in secondary characters, and the appearance of “hallucinated” elements not present in the original painting.
A significant part of the work therefore consisted in developing compensatory strategies to maintain visual coherence and iconographic respect within the limits of the technology.
An Ongoing Exploration
This project does not aim to offer a definitive "animated version" of the works, but to open a space of inquiry between conservation, interpretation, and technological experimentation. The museum becomes a place where the digital does not replace the physical, but accompanies and enhances the direct viewing experience, suggesting new layers of meaning without overriding the original intent of the work.
On these pages, you will find the animations, images of the original paintings, and insights into the prompts used for each piece. I have also included a dedicated section for errors and discarded results, as these, too, are an essential part of the research process.
AI Generated Animations
This section gathers the animations created using artificial intelligence, based on selected works from the Luzzetti Collection.
Each animation is the result of visual analysis, iconographic research, and carefully written prompts, with the goal of preserving the style, composition, and original meaning of each artwork.
The animations are not meant to replace the originals, but rather to offer a dynamic perspective that invites curiosity and opens new interpretive paths—especially for younger audiences.