This project is an homage to the Impressionist movement.
Impressionists sought to depict the world not with rigid realism but with an interpretation that felt truer to human experience.
When the movement emerged in the late 19th century, academic art institutions dismissed it as unfinished or crude. Painters like Monet, Renoir, and Degas had to fight for legitimacy of their art, much like how art that involves generative AI today faces skepticism and resistance from traditional art circles.
These Impressionist artists, as well as Sisley and Pissarro, didn’t invent their style in isolation, nor as soon as they began experimenting. Rather, they observed and imitated one another’s work, borrowed ideas about color, light, and brushwork from each other in a constant mimesis—absorbing techniques and evolving them into new, individualized expressions.
Poplars on the Epte, 1891, Claude Monet
https://www.tate.org.uk/art/artworks/monet-poplars-on-the-epte-n04183
The parallels are strong: critics argue that AI-generated works lack true authorship and creative intent. Oddly enough, this was the same critique that Impressionists had of the contemporary art establishments of their time. And I don’t disagree that most of the AI-generated content presented, a bit forcefully as art, mostly lacks any sense of vision of a singular individual, of the sharing of a unique viewpoint on the world… For that very reason, I had to find a way of using AI combined with a living perspective before I could share it as art, and I think that I have found this through experimentation for this project.
By using AI to reinterpret reality through an Impressionist lens, this project acts as both a tribute and a challenge to the notions of artistic legitimacy, of adherence versus pushback on stylistic formulas, and of representations of human perception through art.
“En plein air” AI video production
… my decision was to take the creative process outside […], engaging with the environment more directly…
I was pleased with this particular approach because it subverts the typical AI workflow. Most AI-driven animations are confined to a fully digital workflow—models are trained on digital datasets, image or video generations are made through a computer screen, and perhaps edited and refined by a human with software in the aim of being published online. My decision was to take the creative process outside of that context, engaging with the environment more directly, and mirroring the Impressionists’ unconventional (at the time) commitment to painting outdoors, or “en plein air”. Their preference was to capture fleeting moments of their real world, and not an idealized setting by composing their frame in their studio. Opting to take the challenge of uncontrolled light, subject and atmosphere, that approach brought a unique quality to the paintings that ensued.
Filming in a way that also embraces spontaneity, my process began on a path in the Mont-Royal park. Each time that I saw something which could be nicely framed, and that had a slight movement of the leaves and branches of trees, I would capture it. Since there were no subjects in my shots, and as a lone-cameraman, the composition of frames happened completely on-the-spot. This approach turned out highly compatible with a human perception-centered approach to creative practice, rather than ideation purely in the form of words, or visually imagining a desired result (like in scriptwriting, or storyboarding). If it was there in front of me, and looked nice, I would film it. Otherwise, it wasn’t going to be part of the piece.
The choice of the Impressionist style
AI-driven style transfer can apply virtually any aesthetic to a given image, but Impressionism felt particularly relevant because of its focus on perception over objective representation. While traditional realism seeks to depict the world as it is, Impressionism captures the fleeting impressions of a scene—the interplay of light, movement, and atmosphere as experienced by the observer.
This aligns with how (re-)processing images through AI works: it “sees” and then redraws pictures of the world. AI does not replicate reality, but mimics it. It interprets and transforms input data based on learned patterns, much like an artist interpreting a landscape through their eyes, their style, habits, and finally through their brushstrokes. This project, therefore, is not just about applying a historical style but about exploring how both Impressionist painters and AI engage in the act of reinterpretation.
Impressionist techniques—loose, visible brushwork, vibrant color contrasts, and the breakdown of form into light and shadow—lend themselves well to the way AI generates images. AI tools like the one used here somehow redraw what is given while simultaneously “understanding” what is “seen”, and being faithful to the perceived appearance of it. The slightly imperfect and unpredictable quality that can emerge from AI processing also echoes the way Impressionists experimented with paint in an abandonment of sharp and detailed lines which ultimately led to their distinctive style.
Alfred Sisley, View of the Canal Saint-Martin, 1870
https://en.wikipedia.org/wiki/Impressionism#/media/File:Alfred_Sisley_001.jpg
Pioneering movements: breaking artistic conventions through rebellion and reinvention
Impressionism wasn’t just a new style—it was a revolution against the rigid academic traditions of the time. Before the Impressionists, the dominant form of art, in France for instance, was dictated by institutions like the Académie des Beaux-Arts, which favored historical, mythological, and religious subjects, painted with meticulous realism and smooth, blended brushwork. Artists were expected to adhere to strict rules regarding composition, color, and technique.
Impressionists challenged all of this by painting contemporary life rather than historical or religious subjects. They didnt stop there: using visible brushstrokes instead of smooth, polished surfaces, they were chasing light and movement instead of static detail. Putting aside intellectual tricks for achieving “correctness”, like accurate perspective and form, they attempted to absorb the visual stimuli with as little preconceptions as possible; without letting the tendency to project their understanding into the world intervene, and favoring an overall more fluid, direct, and perceptual approach.
Their work was initially dismissed as unfinished, amateurish, and even offensive to artistic standards. However, the willingness of the group of artists to break conventions ultimately reshaped the art world, paving the way for modernism and abstraction.
I hope this historical context strengthens the parallel between Impressionism and AI art today. Art presented as making use of AI-generation immediately faces skepticism, just as Impressionist paintings did. The argument that many critics have is that AI lacks intent or true artistic agency, and it currently mostly does, but as a tool for an artist, it can still channel true artistic vision and voice into works of art in my opinion.
Just like Impressionism, looking back at AI art in the future, it will likely represent yet another shift in how we think about creation—not as a static, rule-bound process, but as something fluid, evolving, and for which the public has to be open to new ways of seeing.
Breakdown of the creative process
Walking through Parc Mont-Royal (in Montréal, Québec), I was only equipped with my cellphone camera, and attempted to capture as many “scenes” of the park as I could find.
The many (many) shots I took were mostly static (from experience, I knew that a slight camera motion would give an interesting effect once processed). They then had to be sorted, selected, trimmed, and edited together in a somewhat coherent “story”.
Next, I applied some (very slight) color correction, reframed some shots, added digital push-ins to pull the viewer in, and increasingly applied an animated “liquid distortion” effect on each successive shot to create a hypnotic illusion and accentuate the gentle swaying of the trees.
For those familiar with AI image-to-image processing, my particular approach for AI animation is to opt for quite a low “denoising strength”. For those who aren’t, this is the parameter that dictates how far from the given input image the AI begins its “redrawing”. It is the trick that allows both the adherence to the original images and keeps them “recognizable”. Importantly, since I am doing a frame-based processing (in which solely the current frame’s information is used for the “generation”, or rather transformation of that frame alone), the attenuation of the transformation enables the maintaining of the “moving-image” illusion of video, or in fact animation.
In addition to that, since it uses a fixed seed as well, it further conforms the individual frames to one another and enhances the flow of the images in motion.
And because of the unpredictability of the AI output, I had to re-edit the sequence of paintings/shots so that the overall piece flowed more smoothly. As for the soundtrack, I added a classical music piece contemporaneous to the Impressionist movement, an old personal favourite of mine. I felt like it conveyed the contemplative softness of a stroll through the park, and of the unhurried and perpetual swaying of foliage (drowned, almost, in reverberation and other effects to hint at a nostalgia no longer remembered by living humans).
Serendipity: control, or lack thereof
As a final thought, one of the most fascinating aspects of AI-generated art is the tension between control and unpredictability. In fact, Impressionist painters as well embraced spontaneity—capturing fleeting moments, allowing brushstrokes to flow naturally, and probably being forced to welcome happy accidents in their work.
How much to lean into the unpredictability?
Reflecting on it now, every part of this project has involved surrendering to serendipity:
“On the terrain”
The “en plein air” approach introduces unpredictable elements not encountered filming in the controlled environment of a studio:
- Changing light conditions, shifting shadows, clouds, etc.
- Unexpected movements in the scene (wind, birds, cars, etc.)
- Variations in the objects of a scene, camera angles, framing, and motion
These elements created a dynamic, living quality to the captured footage, much like how Impressionist painters were chasing the ever-changing light and atmosphere.
In post-production
Secondly, post-production was, both before and after the AI step, one area where the balance between curation and chance came up.
What parts of the footage might lead to the most interesting re-interpretation?
How much or how little to intervene in the AI’s output to fit a certain vision and keep control over the result?
How much to lean into the unpredictability and let the AI’s unique “hand” shape the final piece?
In AI interpretations
Crucially for this project, the AI-driven style transfer was not something I could really anticipate:
The AI was emphasizing unexpected textures or colors, similar to how an Impressionist might exaggerate the golden glow of a sunset or the texture of tree leaves.
Details that seemed key to me in my original framing and composition of my shots got frequently blurred or abstracted. The AI’s brushstroke-like patterns were certainly interacting in surprising ways with the natural elements of the base footage.
This entire process consisted of a give-and-take: while I could guide the process, I had to embrace the unexpected, to both make use of and convey the fascinating reinterpretation abilities of AI image-to-image models.
References, and more…
project page: Living Paintings I (for more visual content)
a nod to Davide Quayola
See more of Quayola’s work here: https://quayola.com/selected-landscape-paintings/
PBS’s Art Assignment on Impressionism
*note that Vincent Van Gogh is mostly associated with post-impressionism rather than Impressionism.
David Lynch – Oh! A moving painting!