As I prepared to enter the venue for The Runway 3rd Annual International AI Film Fest, I found myself unsure of what to expect. Around this time last year, I emerged from the Lower East Side’s Metrograph theater let down. I had just experienced the second edition of the fest and proceeded to type up a milquetoast review for our weekly newsletter.

It was hard not to compare it to the prior year’s debut. The first edition had the spark of the new, and a palpable excitement suffused the room. We had already featured our first AI-assisted official selection on Short of the Week (Generation, which picked up the top prize that night), ChatGPT had launched late the previous year, and the infamous NyTimes Bing story landed a few months before the event. AI hype was budding, and the sheer miraculous nature of the tech overwhelmed any skepticism about the present state of the content.

So, last year I thought we might see a leap. But the selections largely disappointed. The best pieces were not primarily generative, and though they leaned into the quirks of the technology in exciting fresh directions rather than blandly replicating existing styles, the results were often chaotic, impressionistic, and perhaps as a result, pretentious and navel-gazing. The tone of the event, while still buoyant, had adopted a defensive posture, as negative sentiment had loudly coalesced online in creative and fan communities, and the now-familiar criticisms surrounding the theft of training data, the specter of job loss, and the philosophical invalidity of machine art percolated.

I’ve been at this long enough to have seen my fair share of hype cycles—I remember when high-schoolers with camera phones were going to be the new Spielbergs, when Interactive filmmaking would achieve a new paradigm of user-agency in storytelling, and when VR was the ascendent medium of the future. I enthusiastically cheerlead some of these, only to be disappointed. I was confident that AI would be different—the breadth of use cases outside of creativity—writing, research, and productivity—ensured a much larger market, and the sheer scale of investment by Big Tech signalled staying power. Yet I wondered if we’d sped run the cycle and entered the “trough of disillusionment” already. AI video would arrive in force eventually, but how long would it take? Video felt like an exponential step up in complexity from text. Would all this hype end up being a decade early, like it seems VR was?

Welp, so much for that. From that moment of doubt onward, AI Video entered an arms race. OpenAI teased their video generation model, SORA, to the public that summer before releasing in the winter, Google got serious with its Veo family of models, Chinese companies like KlingAI burst out of nowhere to become serious players, and the Ghibli-trend saw AI achieve a breakthrough cultural moment. In the past year, Hollywood deals have been struck, illustrious figures like James Cameron have gotten on board, and Natasha Lyonne plans to have an AI feature in theaters soon. A week doesn’t go by without a major announcement, and a whole subculture of AI Creative Influencers has popped up, thriving on X and Reddit. I have yet to really see AI be the primary engine of sustainable and artistically valid practices or models, but FB and TIkTok are suffused with low-rent AI engagement-farming, so higher ambition breakthroughs feel imminent—New York Magazine just this week declared that, “Everyone is Already Using AI (And Hiding It).”

This was the context I brought into last night’s Runway AIFF, and if you’ve come this far into the article and are still looking for a tl;dr, the answer was that this time I was impressed. The event found the venture startup in a form befitting its recent $3B valuation—confident and ready to show off. Importantly, the content had also grown to match this ambition.

[caption id="attachment_41033" align="aligncenter" width="720"] sdds

The scene from inside the venue at Lincoln Center.[/caption]

First, a flex—the festival abandoned the hip and intimate Metrograph to instead take over Lincoln Center. The screening took place at Alice Tully Hall, one of the holiest venues of New York film culture, where acclaimed Oscar contenders debut in the Fall at the New York Film Festival. Signage covered the exterior of the building, and the much larger capacity did not pose a problem, as the event was packed.

I popped in early to grab a drink and survey the crowd. I quickly ran into familiar faces—Cutter Hodierne, best known for winning the Grand Jury Prize at Sundance for his short, Fishing Without Nets, was there alongside Borscht Corp co-founder Andrew Hevia, who now serves as Head of Film & TV for Pablo Larrain’s company, Fabula Productions. Cutter has been using AI for previz on his latest feature, and Runway was partnering with him. I also met Riccardo Fusetti, whose aforementioned short, Generation, won the inaugural fest. He was back as one of the ten finalists in the evening's program with Editorial, which premiered on Directors Notes last month.

We headed into the auditorium for the screening. The two prior editions had been preceded by panel discussions on the promise of AI, featuring figures like Darren Aronosksy. Last year’s edition in particular had been heavily frontloaded with ancillary content full of techworld-style philosophizing about AI, its potential for the creative process, and justifications for its use. While interesting, it also betrayed a degree of insecurity in the face of criticism. This year, perhaps bolstered by the relative strength of the films and the ominpresence of AI discourse in the culture, Runway CEO Cristóbal Valenzuela kept the prelude to a minimum, though proudly noting in his opening remarks that submissions had swelled from 300 in 2023 to 6000 this year, encapsulating the growth of the event and, by implication, the creator community behind it. He then brought on the artist, Flying Lotus, for an informal 12-minute chat.

[caption id="attachment_41034" align="aligncenter" width="720"]

Runway CEO Cris Valenzuela and Flying Lotus.[/caption]

Flying Lotus, the stage name for Stephen Ellison, is an artist we have a lot of time for, so it is worth spending a moment on what he said. The multi-media creator is familiar to us both for his musical collaborations on S/W-selected short films like Until the Quiet Comes, Fuckkkyouuu, and Black Holes, but also for his own burgeoning directing career, such as this year’s celebrated sci-fi horror feature, Ash. His chat onstage and my conversation with him at the afterparty established his excitement for AI tech, which he has actively incorporated into his creative practice.

Ellison demonstrates an iconoclastic streak that is perhaps befitting his unusually varied career, repeatedly espousing his opposition to “rules and gatekeepers”. He is firmly in the “it’s just a tool” camp, and scoffed at blanket dismissals of AI from critics, noting for the audience that when coming up in the early 2000s he got “a lot of flack” because he made music on the computer, repeatedly being told, “You can’t make hip-hop on no computer, what the fuck is that?” he recounted to laughter from the crowd.

"It becomes a conversation with the medium, which is what’s so new and interesting"

Ellison says that he primarily uses generative AI in development, and that advances in the tools have surprised him in “how much easier it’s getting to convey the ideas, and how much easier it is to wake up with an idea, or NOT to wake up with an idea which is then the real fun part, the fuck around and find method, which AI is really good for.” He went on to describe the novelty of the collaborative process, a sentiment that I’ve heard from many using AI, saying, “…you don’t always get what you wanted, and then maybe you pivot what your idea was and go into a different tangent—it becomes something that you didn’t intend and it becomes a conversation with the medium, which is what’s so new and interesting about this stuff.”

While Ellison expressed empathy for those who are pessimistic about the implications of AI, his enthusiasm for the tech remains undimmed, and the biggest applause line came when he answered the hypothetical, “Aren’t you afraid that people can just one day, prompt a Flying Lotus song?” His response, “Well motherfucker, I’ve got the tools too! I got the tool too, so now y’all really in trouble!”

[caption id="attachment_41035" align="aligncenter" width="720"]

Scene of the audience from the screening. 📸 Runway.[/caption]

The lights dimmed shortly after, and the program, comprised of 10 finalists averaging around 5min apiece, began. If I hadn’t stayed abreast of advances in AI video in this intervening year, the first piece would have shocked me. More Tears Than Harm is a brightly styled animated portrait of Madagascar, experimentally comprised of fleeting images like a travelogue, but with a mature visual style and consistent aesthetic. It’s not groundbreaking (I would be shocked if the artist was unfamiliar with Bastien Dubois’ 2012 short Madagascar, carnet du voyage), but legitimately impressive in how it realizes an attractive fine art animation look that didn't feel derivative.

Another film with a distinct animated style that would have looked at home at Annecy was the Ukrainian short, Distance Between Two Points of Me, which delves into a meditative discourse on presence and absence, influenced by the devastation that war in the country has wrought.

As that description suggests, much of the program was pretty heavy, including films about a child born with a fatal birth defect and chickens rescued from factory farming. The selected films also suggest that AI filmmaking still has a slightly obnoxious tendency to engage philosophically with the means of its own creation. Fusetti’s piece was therefore a welcome respite, with its horror twist and deadpan punchline ending providing needed variation in tone to the evening. Maddie Hong’s Emergence, a first-person journey of a cicada emerging from its long periodic cycle, plays into my complaints of grandiose faux-profundity, but was excellently executed, and welcome for the novelty of its perspective.

That said, my favorite of the program and the Jury’s ultimate Grand Prix winner was possibly the most pretentious of the bunch. Jacob Adler’s Total Pixel Space is a slow, almost scientific exploration of a single remarkable idea, that the nature of digital visual information means that every image that has or could ever be is, theoretically, finite, waiting less to be designed or captured, but instead discovered. Visually, it results in an excuse to throw a collection of oddball images onto the screen, which AI excels at, but the sober presentation of its voiceover and the sheer improbability of how far Adler stretches the idea over 9 minutes achieves a magical, almost hypnotic effect on viewers.

I caught up with Adler afterward, and he is an interesting model of an artist empowered by AI. A musician by training, the idea for the short had been brewing in his mind for several years, but he didn’t know how to express it. AI gave him the language to communicate his ideas, and the pleasure he’s experienced creating AI videos has inspired him to dive headfirst into the craft of non-AI filmmaking.

As this quick summation suggests, the quality of this year’s finalists was a marked leap from previous editions of the festival. Some of it is the tech–consistency is finally here, which makes a huge difference, and the adaptability of cutting-edge models to a diversity of visual references, from hyperrealism to bespoke animation styles, was readily on display. But some of it is the storytelling, too. There was a concision and clarity to many of the ideas in these shorts that contrast with the selections of prior years. It also contrasts with traditional shorts, which we've noticed to be swelling in length in recent years. That none of the ten finalists cracked 10 minutes was an interesting thing to note.

Now it is our turn to be pretentious—if you're asking how good the films are in absolute terms, I don’t think any of these films will make the cut on S/W. But, there is a legitimate conversation to be had on several, and I look forward to sharing them with my team.

***

The close of the evening brought my thoughts back to Runway, which, at no small expense, put on the event and will host its counterpart in LA at the Broad Stage Theater next week. It is interesting to consider the motivations behind taking an art form that is intrinsically digital, and whose community is primarily online, to an offline event. Some of it is surely validation, borrowing the style and romanticism of film festivals to impart prestige upon the nascent medium, some of it is community-building, and a large chunk of it is, of course, marketing. But kudos to Runway for not restricting applicants to their own tech. Despite being in a heated race with companies that have essentially infinite pocketbooks like Google and OpenAI, many of the finalists did not utilize Runway’s ballyhooed Gen-4 model. Runway’s name was on all the signage, but the event was not a product showcase, and instead a more general and inclusive celebration of the emerging form.

That speaks to a true-believer quality the company seems to possess. I have no illusions that a sizeable portion of our audience hopes that all these companies crash and burn, but if one were to have a rooting interest in the tech wars to come, Runway is a sympathetic underdog. Co-founded by Valenzuela in 2018 out of NYU’s celebrated ITP Lab, a place sometimes referred to as “art school for engineers”, the company was prescient in how early it was to the potential of generative AI, and this origin story embeds an artist-first mentality into Runway’s DNA. Other companies have imported artists into the fold, such as OpenAI with Blumhouse and its filmmaker incubators, or the flurry of partnerships Google announced last month with the likes of Aronofsky, or James Cameron being added to the board of StabilityAI. Our friend Paul Trillo has even taken a senior role at Asteria. But tech history has, for good and ill, demonstrated the power of a founder in aligning a company’s strategy to its vision, and as the film festival demonstrates, I don’t think it’s an accident that Runway’s artist partnerships and community-building initiatives are more advanced than their competitors and seemingly more organic. Valenzuela may not be as charismatic as a Steve Jobs, but he evinces sincerity in the storytelling ideals his company espouses, as seen in his recent interview with The Verge, and that is a quality that will be difficult for these megacorporations to match.

[caption id="attachment_41037" align="aligncenter" width="480"] Heavia and Fusetti on the way to the afterparty.

Heavia and Fusetti on the way to the afterparty.

Hevia and Fusetti, on the way to the afterparty.[/caption]

If pressed for an overarching takeaway from the night, I will reiterate a sentiment that loyal readers of the newsletter will recognize—it's not there yet, but it feels close. For detractors, my apologies, but there doesn't seem like any stopping this train, nor do I think any coordinated taboo seems likely to hold. As the tools are enthusiastically, or begrudgingly, incorporated, I wonder if the age of AI films attracting attention for being AI, for the novelty of that designation, and the inherent curve we judge them on may soon be put behind us. The long-term viability of AI Festivals like Runway’s may dissipate as films become absorbed into the existing infrastructure. Established festivals around the world are affirming the eligibility of AI-assisted works as selections, and though in practical terms it only affects a few, do not underestimate what a big deal it was that The Academy clarified that AI works will be eligible for Oscar consideration going forward.

I can’t pretend to know how AI will change the financial structures of the larger film industry, or how projects are made, or even what positive or deleterious effects AI may have on the form itself. But walking through Central Park last night on the way to the subway, I was struck by the sense that very soon the question we ask of an AI film may be the same we ask of every other film as the images fade and the credits roll—was what I just watched great?