March 2025
VMAC Article
The Omnipotentiality of AI Films?
Text: Phoebe Wong
Last November (2024), I spent two weeks in the Netherlands and attended the very popular documentary festival, the International Documentary Festival Amsterdam (IDFA). This festival is of interest to me because of its DocLab, which focuses on digital storytelling and XR with a documentary impulse.
DocLab turned 18 this year. While a plethora of VR and AR projects were expected, I also encountered many AI-driven works. A third of the 27 showcased pieces in the DocLab programme employed AI as a tool, while full-length documentaries appeared in the main festival events. When you think about it, why wouldn’t they? AI has become ubiquitous. There are even dedicated events for AI films, such as Runway’s AI short-film competition and the Flux Festival in Los Angeles, which debuted in November 2024 as a festival exclusively for AI-driven cinema.
As AI artworks gain momentum, what kind of AI aesthetics are emerging? The curator of the Flux Festival analysed the showcased works on display and identified five prominent trends in AI-generated imagery: 1. Morphing, blurring visuals; 2. surrealism; 3. Dark narratives; 4. Nostalgia; 5. new times, new space. [1] These trends are also evident among the AI works presented at the IDFA.
These works – Emeline Courcier’s Burn from Absence, Tamara Shogaolu’s Oryza: Healing Ground, and Gregor Petrikovic’s Sincerely, Victor Pike – use AI to uncover or fill in the gaps, whether in hazy memories or deliberate forgetting, but with their own twists.
Emeline Courcier, a Paris-born artist, explores her Vietnamese roots in her new work, Burn from Absence. The nostalgic film, presented as a video installation, reconstructs her fragmented family history. It combines old family photographs, AI-generated images based on those photographs—I could not tell which were the originals—and stories of her elders fleeing the ravages of war. As the AI reassembles the broken and fragmented memories, it also blurs the line between reality and her recollections.
Ephemeral as dreams, fleeting as illusions. Artist Gregor Petrikovic, driven by his faltering memory, begins recording conversations with those around him. His short film Sincerely, Victor Pike delves into the fragility of memory, portraying it through a stream of anomalous and blurred images. Using AI-generated visuals derived from his voice recordings, cars take flight, and bodies intertwine. As Petrikovic observes, “This may actually be true to the nature of memories, which lie on the boundary between the real and the surreal.”
As for using AI to craft mysterious, dark tales, About a Hero is one such example. It is also the opening film of the Festival.
“A computer will not make a film as good as mine in 4500 years.” This bold claim by legendary filmmaker Werner Herzog sparked Polish director Piotr Winiewicz’s curiosity: could a computer write a film script? According to Winiewicz, when he began conceptualizing this project in 2018, AI was not as ‘readily available’ as it is today. He had to develop a programme named ‘Kaspar,’ an AI modeled on Herzog’s ‘mind’—emulating Herzog’s way of thinking, language, and cinematic style—to craft the script.
About a Hero was thus born, featuring Herzog’s ‘voice performance’—an AI-synthesized voice-over that, with his approval, serves as a narrator, an investigator within the plot, and a skeptic of AI. The story begins with a lawyer explaining the capabilities of deepfake technology, establishing the film’s dialectical tone that explores authenticity and deception.
Winiewicz calls About a Hero an essay film, albeit disguised as a drama, centered around the investigation of a case in an imaginary industrial town. The suspenseful plot serves as a ‘draw.’ The fragmented storyline—whether intentional or a byproduct of AI screenwriting—remains ambiguous, yet it undeniably heightens the tension of the mysterious case. The performances, from the widow of the deceased to the town’s neighbors and friends, evoke Roy Andersson’s signature cold absurdity. Notably, the film’s visual aesthetic is reminiscent of the hyperrealistic work of German sculptor and photographer Thomas Demand, where the pristine cleanliness adds to the eerie atmosphere.
The director’s use of talking heads, a conventional documentary tool, demonstrates restraint, as it keeps the film accessible and prevents it from becoming overly complex. Three types of talking heads are deployed in the film: first, the witness interviews that reveal traces of exaggerated fictions; second, the found footage of art theorist Boris Groys, media artist Stephanie Dinkins and other discussants ‘grafted’ onto the investigation’s narrative, adding a touch of pseudo-documentary; and third, the media lawyer (who discusses and ‘demonstrates’ deepfake) and cultural luminary Stephen Fry ponder the ethical and philosophical implications of AI. The drama overlays three expressive dimensions—fiction, appropriation, and contemplation—intertwining truth and falsehood, for an eclectic essay film. Among them, appropriation stands out: it critiques the authority of documentary-style talking heads while exploring the mysterious ‘machine,’ naturally leading audiences to reflect on AI.
About a Hero captivates audiences with its hybrid style, embracing complexity through a dialectical, and reflective stance on the AI—likely the reason it was selected to open this year’s IDFA. In contrast, Eno, a promising generative documentary, received comparatively less attention at this Festival, not to mention a very conventional screening method of two showings in the cinema. (Eno was recently organised as a 24-hour globalised event, so audiences could have chosen to watch different versions of the film one after the other.)
Drawing on 30 hours of interviews and 500 hours of material, Eno is a cinematic portrait of the maverick musician, composer and artist Brian Eno. For the documentary, filmmaker Gary Hustwit and digital artist Brendan Dawes have developed a generative software, which combines Eno’s interviews with his archive of unreleased video footage and music, resulting in an infinite number of permutations. [2] Stochastic, fun. Each screening is set to begin and end in the same way, leaving the rest of the film to unfold unpredictably. It explores not his personal life per se, but his overarching thoughts on music and creativity.
Aside from being a unique feature-length production and follows loosely to the narrative structure of documentary, the resulting generative film is intriguing in the following ways. Firstly, it feels like an unfolding of the filmmaking process, which is known to be inherently disjointed and fragmentary; it is only during the later stages of editing that things come together and make sense. Secondly, it plays with the dialectic of cohesion and coherence, exploring the space in-between. Finally, to borrow a term often used and cherished by Brian Eno, “surrender,” the brave new world of generative film (and documentary) relies on a willingness to surrender—to relinquish control and allow spontaneity to take over. The film is a vivid example of this.
From the works mentioned above, we glimpse the ‘omnipotential’ of AI. While consumer-grade AI has transformed algorithmic art into what we now call AI art, generative films are actually nothing new. It also remains apparent that the most sophisticated creations still necessitate programming expertise.
Generative art is like a river: one can never step into the same river twice. The river of images flows ceaselessly, therefore it is theoretically an infinite possibility of combinations. Some of my favourite works from the early-2010s include Linda Lai’s generative video series Door Games Window Frames and Ian Cheng’s live simulated animations.
Take Linda Lai’s Door Games Window Frames: Near Drama (2012) as an example. Lai selected clips from the old Cantonese movies from the 1950s and 60s and cut them into basic units of 2-5 seconds in length. With a total of more than 500 units from 11 films, they are stochastically combined and remixed by a computer program coded by Lai. The resulting video randomly assembles vaguely familiar fragments, creating a journey from the known (media footage) to the unknown.
In his 2023 essay ChatGPT Is a Blurry JPEG of the Web, sci-fi novelist Ted Chiang compares large language models (LLMs) like ChatGPT to lossy compression algorithms such as JPEGs. He argues that while these models retain much of the information from their training data, they offer approximations instead of precise reproductions. Chiang highlights the limitations and potential distortions of such AI systems, and how AI models, being programmed to paraphrase, may give the impression that they are ‘thinking’ and thus appearing intelligent.
Chiang’s analogy offers a nuanced critique of AI, likening it to a lossy compression of human knowledge, capturing the essence but losing clarity—what a thought-provoking take. To follow Chiang’s line, how would these ‘blurry JPEGs’ lead to a new form of (virtual) storytelling?
[1] See the The Conversation article written by Holly Willis, Professor of Cinematic Arts at the University of Southern California: https://theconversation.com/blurry-morphing-and-surreal-a-new-ai-aesthetic-is-emerging-in-film-242098.
[2] According to The New York Times, there are 52 quintillion possible versions of Eno. (1 quintillion = 1,000,000,000,000,000,000.)
(The views and opinions expressed in this article published are those of the author/s. They do not necessarily represent the views of VMAC.)
Staff Pick
The Book of Hong Kong Independent Video Competition 1993
In April 1993, the Hong Kong Arts Centre organized the first Hong Kong Independent Video Awards, at which Videotage presented an additional screening programme entitled ‘Poetry and Thoughts’. Its curator, May Fung, was also one of the judges in the ‘Experimental Video’ category of the competition. Earlier that year, the Fringe Festival presented a synonymous programme [1], whose theme encouraged artists to explain their understanding of ‘video’ beyond their works.
The competition’s official booklet not only lists the winning videos, but also the details for the screening programme, which includes the artists’ reflections alongside the synopses of their titles to be presented, such as Ellen Pau’s Song of the Goddess. In her article, Pau discusses how, without long-term arts policies and government subsidies, independent creation is limited by the approval of the officials and sponsors despite all the idealistic possibilities. Between the themes and resources that such can attract, artists cannot freely tell non-mainstream stories, affecting their freedom of creation and speech.
Two years later, the Hong Kong Independent Video Awards merged with the Independent Short Film Competition, organized by the Urban Council, to form the ifva Awards [2], which has become a cradle for nurturing Hong Kong’s new generation of creators. In September 2024, the latter announced that it will be taking “a brief hiatus” [3], which could be a prelude to the end of an era?
Pau ended her article by describing Hong Kong’s creative scene in 1993 as “winter”. Today, government grants are still a main way for artists to get funding, but the current socio-political atmosphere has made it harder to express one’s thoughts, in addition to the dwindling number of ways to make one’s mark. What would the city’s independent creation look like in the future?
[1] Compared to the ephemera collected in Asia Art Archive’s ‘Ellen Pau Archive’, although both programmes shared the same title, the lineup of videos presented were slightly different. See https://aaa.org.hk/archive/309205 for the Fringe ‘93 version of the programme in January.
[2] Sourced from ifva’s interview with Jimmy Choi and Lo Tak-sing in 2010.
https://web.archive.org/web/20110301083750/http://www.ifva.com/2010/page_en/about/birth.php
[3] https://ifva.com/page/?id=3WLjc6yOQOM
Video Review
Suck/Blow by Leung Chi Wo, 2003
Selected and reviewed by Candace Cheung Wai Chu (Intern)
How long has it been since you last emptied yourself, and felt your breath calmly? The 29-second blank at the beginning of Suck/Blow can be a transition for you to focus on your breathing.
Leung Chi Wo uses monochrome photos, gradually intensifying the sounds of inhalation and exhalation and suddenly zooming in and out to anthropomorphize the imagery. This creates visual impact and engages our senses, allowing the audience to fully ‘feel’ in addition to seeing. Throughout the video, no character speaks or stares at the audience, yet interacts with them through synchronized breathing, turning viewing into natural performance art.
The artist offers an interpretation of the form of ‘video’: video is the image in motion. The structure of the video is thought-provoking: a single photograph is zoomed in and out, then two, then more, and finally back to a single image. It’s like the breathing, in and out, of one person, then two, then many, and finally ‘we’ as a whole. We wish to feel comfortably detached from our surroundings, but often find ourselves involved. Just as the corners of the frame become darker as the images are superimposed, we become more attached to the scene as our breathing involuntarily moves in the rhythm of the video.
The buildings frame and define the shape of the sky as we look up, making us feel inextricably linked to the city. When we inhale deeply (zoom in), everything seems empty, and when we exhale fully (zoom out), the edges of the buildings resurface. The cross-shaped sky appears to represent redemption, but we actually forget that it is actually shaped by the annoying skyscrapers. White cannot be defined without ‘black.’ So the structures exist in a grey area, entirely dependent on how the viewer sees them.
As the overlapping sounds of breathing intensify and multiple photographs are superimposed, inhaling and exhaling become indistinguishable. The city exists as a kind of organ and we, standing between the buildings, are the heart of the city. We do not need to draw borders between the city and beyond because we all live under the same sky.
About VMAC Newsletter
VMAC, Videotage’s collection of video and media arts, is a witness to the development of video and media culture in Hong Kong over the past 35 years. Featuring artists from varied backgrounds, VMAC covers diverse genres including shorts, video essays, experimental films and animations. VMAC Newsletter, published on a bi-monthly basis, provides an up-to-date conversation on media arts and their preservation while highlighting the collection and its contextual materials.