Will Generative AI Kill Slow Cinema?

One of my favorite movements in film is slow cinema. Those delicious, often ridiculously long takes are almost therapeutic and perhaps an antithesis to the fast, hyper-cuts that define today’s blockbusters. You can even see its influence in commercials, music videos, and indie films too.

Recently, I finally sat down with Jeanne Dielman, 23 quai du Commerce, 1080 Bruxelles (1975), directed by Chantal Akerman, and let its unhurried rhythm wash over me. In one scene, in particular, the camera holds absolutely steady as she prepares a meal. In real time you watch the event unfold.

It’s this deliberate pace — the invitation to linger, to let your mind wander (and boy, do you have time) — that gives slow cinema its unique emotional gravity. This added depth evokes a sensation that might follow studying a painting carefully for twenty minutes; it’s that exploration and sense of place and time that makes for a unique cinematic experience, and explains why I really enjoy these slow cinema experiments. After nearly three and a half hours, a runtime that exceeds even films like The Godfather (1972), I understood exactly why Jeanne Dielman keeps popping up on “best films of all time” lists.

Of course, Akerman is but one example of a filmmaker daring enough to push the envelope. Many other names come to mind: Tarkovsky (Stalker, 1979), Tarr (Satantango, 1994), and maybe on certain occasions modern auteurs like David Lynch (Twin Peaks: The Return, 2017, for instance).

Meantime, I’ve fully jumped into the AI rabbit hole. I’ve found a few killer apps. One is using these AI bots as server admins to help run this site. What a relief not to have to venture into the IT Dungeon alone. ChatGPT, Claude and others (Cursor, Gemini) have also been instrumental in speeding up location scouting and brainstorming scene structure and flow for short films and some of our upcoming experimental projects. For example, I needed to arrange dozens of locations in San Francisco and plan an efficient shoot. ChatGPT made short work of that prompt, quickly and effectively providing a list of locations and dates, and a way to minimize the number of trips I’d need to make up 101 to the city so my time was well spent.

Then there’s music and scoring production. AI can really accelerate that as well. Not necessarily creating it all from scratch, beginning to end (though, as we know, that’s possible), rather to act as a studio assistant of sorts. Pulling together fragments of my ideas and giving the structure. Sparking further ideas. Helping me even get started by pulling together a mix of patches that might make sense to use on a virtual Roland D-50. I no longer need to do this entirely alone — occasionally bouncing an idea off Loni — and now feel like I have a team to assist and support my work.

I wondered as I waded through this new world of AI: Will gen AI kill slow cinema as we know it?

We could even broaden the scope and ask how many things it will kill. Or, given it’s rapid advance, maybe it would just be easier to shortlist things that might survive. And in that respect I believe slow cinema will be one of them. (not to mention human art in general, but that’s a conversation for another day).

With that in mind, here’s why I think the AI vs. human outcome will be more akin to a creative partnership, and not a dystopian death match.

Here’s why I believe AI and slow cinema will coexist as collaborators, not competitors:

Algorithms Don’t Have Patience

AI loves efficiency. It’ll pump out a “two-hour” slow-burn in about two seconds, complete with CGI moss on the walls and slightly off-kilter Dutch angles. But like ordering a tiny latte and getting it in a red plastic cup, you sort of lose the ritual. Slow cinema demands time, not just timecode.

The Human in the Frame

Tarkovsky’s shots linger because each one is an existential question: “Why is that stick floating in a pond?” AI might replicate the movement of the camera, but it can’t ask that question. It doesn’t perceive the stick. It only sees pixel data.

Intent Over Instinct

Lynchian long takes aren’t merely “long” for effect. They’re invitations to drift, to notice a stain on the ceiling, to feel uneasy about fluorescent lighting. AI can mimic motion blur, but it doesn’t understand disquiet. It can’t dream; it can only regurgitate — just as the LLMs (Large Language Models) have been trained to do.

Collaboration, Not Competition

Here’s the upside: AI can handle the grunt work. The bots can test renders, color experiments, variations on “dust mote in sunlight.” That frees directors to focus on the moments that matter: the patient build, the suspended heartbeat, the tiny detail that lingers. Again, an algorithm-driven partnership.

Future-Proofing Slow Cinema

In the end, AI is a tool, not a tyrant (at least not yet). It can’t kill slow cinema any more than Instagram killed quiet mornings. If anything, it might spark a renaissance: filmmakers using AI to refine their long takes, push boundaries, and remind us why we love to linger on the frame.

The Dream State: Why AI Can’t Capture the Unconscious

But there’s more to dig into here. The distinction between mimicry and genuine artistic intent goes deeper than technical capability. When Andrei Tarkovsky held his camera on a glass of milk trembling on a table in Stalker, he wasn’t just recording duration — he was channeling something ineffable through that duration. The trembling glass becomes a meditation on fragility, on the invisible forces that shape our lives, on time itself dissolving into pure experience. Though my interpretation may differ from yours, the essence remains.

True, AI can reproduce the shot perfectly: the exact angle, the precise tremor frequency, even the way light refracts through the milk. Yet it operates from pattern recognition, not from having lived through a moment where a simple glass of milk suddenly contained the entire universe. It lacks what Jung called the collective unconscious — that deep reservoir of human experience that artists tap into, often without fully understanding it themselves. Not simply tap into a scraped LLM, however effective it may be at capturing our lived experiences.

This is why Lynch’s infamous 18-minute sweeping scene in Twin Peaks: The Return works. It’s not just long; it’s inhabited by decades of Lynch’s dreams, anxieties, and obsessions with the mundane becoming sinister. The duration becomes a portal into the director’s psyche. AI has no psyche to share, no dreams to disturb us with, no personal relationship with dust motes or fluorescent hums. It can create uncanny valleys, but not uncanny revelations. Note: Part 8 from Season 3 (the one with the nuclear explosion) is a gift to fans of art-house.

When we watch slow cinema, we’re not just watching time pass—we’re sharing in another human’s perception of time, filtered through their unique consciousness. That’s the intent behind the instinct, the dream within the data.

Slow Cinema Lives: The Movement’s Vital 2020s

Fortunately, we needn’t frame slow cinema as a relic of the past. I believe (hope!) that slow cinema has found renewed purpose in our accelerated, hyper-AI age. The 2020s have produced remarkable works that prove the movement’s continued vitality and evolution.

Apichatpong Weerasethakul’s Memoria (2021) uses extended takes to explore sound, memory, and existence itself. Tilda Swinton’s character experiences a mysterious sonic boom that only she can hear, and Weerasethakul lets us sit with her confusion, using duration to make the invisible audible. The film’s famous final shot — a mesmerizing 20-minute conversation with a man who remembers everything — transforms time into a character itself. This is a film that I think of frequently, and warrants repeated viewings to fully appreciate its message(s).

Tsai Ming-liang’s Days (2020) strips away even dialogue, following two men through their solitary routines before they briefly connect. Shot during the early pandemic, it accidentally became a document of isolation that resonated globally. Its long takes feel less like aesthetic choices and more like survival strategies. How do we move forward in a world that has seemingly stopped?

Kelly Reichardt continues her patient explorations with Showing Up (2022), where the act of making art (sculpting) is shown in real time. The film argues that creation itself is a slow cinema act, requiring presence and duration that no algorithm can compress.

Even younger filmmakers are embracing the form. C.W. Winter and Anders Edström’s The Works and Days (of Tayoko Shiojiri in the Shiotani Basin) (2020) is an 8-hour rural epic that follows a farmer through seasons. Rather than feeling archaic, its duration feels radical, completely rejecting the Netflix culture of binge-speed consumption. So it’s no surprise that this film is hard to find on streaming platforms.

These films don’t just maintain slow cinema’s traditions; they expand them, finding new reasons why dwelling in time matters more than ever. In an era of TikTok and AI-generated content, they’re acts of resistance that insist on human-scale perception, proving that some artistic expressions can only emerge through patience, presence, and lived experience. I’m ever thankful to stumble upon these works of amazing art (thanks Letterboxd for helping me keep all of this sorted and tracked).

AI Tools Broaden the Creative Landscape, Expand Possibilities — But the Human Touch Still Matters

In the end, my money’s on slow cinema. Because it’s not a style you can automate out of existence.

I’ll continue to explore creative ways to incorporate AI into my workflows, especially for film and music (while ChatGPT and Claude take care of the Stark Insider server!). These tools broaden the creative landscape, not threaten it.

After experimenting with gen AI filmmaking tools (e.g. ImagineArt Video Studio, Google AI Studio), I’m more convinced than ever: they excel at speed and efficiency, but they cannot sit still. They cannot hold a shot for seven minutes on an empty hallway and find meaning in the dust particles. They cannot feel the weight of waiting.

When I tried using AI to generate contemplative or experimental sequences, the early results were hollow and lacking. What these tools revealed is that slow cinema (and maybe even filmmaking in general) isn’t just about duration; it’s about the human capacity to be present, to endure, to find the sacred in the monotonous (like broken link checking this site, sigh). A human must still manage the process, not just as curator or technician, but as the consciousness that transforms time into meaning.

Still, to be fair, these new gen AI tools are interesting to try, and many I think will find their way into my workflow. Who can deny the appeal of working faster and more efficiently?

So take a breath, embrace the pause, and yes, maybe even let AI hold your camera (or iPhone) for a minute. Just don’t trust it to find the soul.