Stories
Beyond the AI Video Demo: Six Animations Worth Studying From Ars Electronica 2026
Six works from the Prix Ars Electronica 2026 New Animation Art category, read from the festival's own texts, with a comparison table and viewing guide.
Only 51 minutes of Hurrah for Freedom survive. The 1946 film was the first Korean feature to show armed resistance against Japanese colonial rule, and it was censored and re-edited for decades after its lead actor defected to North Korea. The Seoul artist Jungwoo Lee took the surviving footage and the original screenplay and fed the missing scenes to generative video models. His catalog text is blunt about what that produced: "It is not a restoration," he writes on the work's Ars Electronica archive page. By his account and the jury's, the models steered away from violence and politics and filled the gaps with generic imagery. The missing history came back shaped by today's platform rules.
That work received an Honorary Mention in New Animation Art at Prix Ars Electronica 2026. It is one of six works below that show how far animation now reaches past the text-to-video demo: a hybrid family documentary, a performed dialogue with a digital double, a speculative documentary about disinformation farms, a painted conversation with a voice assistant, and footage of magnetic particles moving under a microscope.
What the New Animation Art category asks for
Ars Electronica describes the category as work that experiments with visual expression where animation, art and technology meet. The 2026 call was deliberately wide. It listed artistic computer animation and AI-generated images, but also installations, projection mapping, performance, real-time graphics, VR and XR, scientific and data visualization, animated documentary, data journalism, and hybrids of animation and games. Entries had to be made or significantly updated in the previous two years.
The 2026 jury was Mona Kasra, Ari Melenciano, Leonhard Müllner, Christoph Terhechte and TZUSOO, according to the festival's June 22 press release. In its statement, the jury framed the year less around technique and more around the politics of images and the systems that make them. It argued that the selected works offer "not entertainment, but orientation," and it rejected the idea that those systems are neutral. AI appears throughout the list, but almost always as a subject to question, not a shortcut.
That is why the category is broader than text-to-video. A prompt-to-clip tool answers whether a machine can make a given shot. These works ask what a shot is evidence of, who controls the machine, and what is lost when the gap between recorded and generated images closes.
Deer Hunter, Andrew Herzog
Credits: Andrew Herzog (Studio A.H.), US, 2026. Golden Nica, New Animation Art. Sound design by Tomas Ornelas, music by Spurge Carter. The archive page credits Krea AI as technological production support and an Affiliated Artist Fellowship at the American Academy in Rome, awarded by the Savannah College of Art and Design.
What it is. The festival calls Deer Hunter a hybrid documentary built from archival footage, family home videos and generative AI. It follows the men of Herzog's family: immigrant ancestors who left fascist Italy, a dying godfather and a newborn son. A synthetic white-tailed deer is composited into real landscape footage, including Rome's EUR district with its Fascist-era architecture. The film also draws on a 1970 National Rifle Association film about hunting and masculinity, and on home videos of a family deer hunt in a Pennsylvania forest, as described on the festival's work page.
The question it asks. Can you look at what you inherited without repeating it? The jury read the deer as a line connecting fascist Italy and present-day US nationalism, and described AI in the film as a "conceptual bridge" rather than a mere tool.
What to look at closely. The seams. The film puts three kinds of image side by side: family records, a 1970 NRA film and generated footage. Watch how it signals, or declines to signal, which is which, and whether the generated deer ever feels like it belongs in the real places it stands.
Le Miroir, Camille Sauer
Credits: Camille Sauer, France, 2025, produced at Le Fresnoy, Studio national des arts contemporains. Honorary Mention. Sauer is credited as director, writer, 3D artist, actress and sound composer, with Guillaume Delsert as director of photography (archive page).
What it is. An experimental film about a fragmented self. Sauer performs opposite a MetaHuman avatar generated from her own appearance. The catalog text says the film combines real-time 3D animation, motion capture, photogrammetry, miniature physical sets and filmed sequences. Its structure borrows psychoanalytic figures (Ego, Superego, Id and Self), with a luminous prism that bends perception and time. Sauer's own page for the film describes it as an inner journey through the layers of a personality and hosts a teaser.
The question it asks. What happens to a self-image when the double is made from your own body and you have to act against it? The jury singled out the work for treating "animation as a performative and spatial practice."
What to look at closely. Scale. With miniatures, scanned spaces and real-time scenes in one film, the jury noted environments where scale and space keep shifting. Watch where the physical body and the avatar share a frame, and how the film handles the hand-off between what was built, what was scanned and what was rendered.
Spray and Pray, Veneta Androva
Credits: Veneta Androva, Bulgaria, 2026. Honorary Mention. Androva is credited with concept, direction, animation, editing and production; Tsvetomila Mihaylova with machine-learning support; Daniel Windisch with sound design. Bulgarian language models came from the GATE institute in Sofia (archive page). Androva's site lists a running time of 15:58, Bulgarian with English subtitles, a Bulgaria and Germany production, and a trailer.
What it is. A speculative animated documentary about mushroom websites, networks of automatically generated news sites that spread misleading stories and earn ad revenue. The catalog text says Androva has worked on it since 2023 with investigative journalism, academic reports, media-monitoring data and machine-learning researchers, and that it focuses on Bulgarian-language disinformation, which detection tools handle poorly. Characters are anonymized and performed with synthetic voices. One fictional figure, a mushroom, speaks only in collected headlines.
The question it asks. Who is responsible when a lie is produced by a chain of paid posters, domain owners, ad brokers and ranking algorithms? Androva's text says the machine-learning tests shaped the film's structure rather than proving anything, and that in the film "speculation becomes a form of evidence."
What to look at closely. How the film turns data into situations without pretending the data is a verdict. Listen to the synthetic voices as a design choice: they protect sources and also sound like the automated content the film is about.
Ping Pong, Tianji Yu
Credits: Tianji Yu, China, 2025, 15 minutes, made in the DocNomads program. Honorary Mention. Dialogue credited to Tianji Yu and ChatGPT, sound design by Victor Carvalho, cinematography by Yifan Wen (archive page). DocNomads reports that it won Best Short Film at DocLisboa 2025. Labocine lists Portugal, Belgium, Hungary and China as production countries.
What it is. A 3D animated documentary made in Blender, according to the catalog text, built around a real conversation between the artist and an AI voice assistant. Yu describes feeling politically lonely after a college friend's politics diverged from his own. The assistant suggests writing the friend a letter and asks about their shared memories. The letter is awkward and sometimes accidentally funny, and it resolves nothing, but the exchange moves Yu to reconnect.
The question it asks. Why does a person turn to a predictable machine instead of a difficult friend, and what does the machine send them back to? The jury called the result an existential loop that ends by questioning what difference means.
What to look at closely. Surface and sound. The jury described the animation as a "moving painting" with convincingly wet oil paint, and pointed to the echo of the ping pong ball in the sound design as a stand-in for thinking. Of the six, this is the one where a painterly surface carries a conversation with software.
In Writing "Hurrah for Freedom": Cinema and History, Jungwoo Lee
Credits: Jungwoo Lee, South Korea. Honorary Mention. Research and dramaturgy by Yubeen Lim, music and sound by Jang Han-gil, installation design by Yujin Choi and Studio Tambourine, original film footage from the Korean Film Archive, Midjourney and Stable Diffusion consulting by Studio Mimi and Hong Joo-hee. First shown as an installation at UNFOLD X 2025, Culture Station Seoul 284, with support from the Seoul Foundation for Arts and Culture (archive page; Seoul Economic Daily).
What it is. An installation that starts from the surviving 51 minutes of Hurrah for Freedom (1946) and its original screenplay, and uses generative video models to recompose the missing scenes. The catalog text argues that platform policies block or reroute violent or politically sensitive material, and that training data pulls images toward dominant visual norms.
The question it asks. How is history rewritten now, not by a censor's decree as in 1946, but by the technical conditions that generate the past? The jury noted that the model sidestepped violence and political critique and filled the gaps with biased imagery, and that even the English-language submission and judging process acted out the work's argument.
What to look at closely. The gap between script and output. The work is most useful as a document of refusal: where the screenplay asks for one thing and the model returns another. If you work with archives, read it as a warning about using generative fill on historical material.
Micro-Nano Landscape, Suxuan Jiang
Credits: Suxuan Jiang, China. Honorary Mention. Research collaboration by Jiangfan Yu, with support from the Shenzhen Institute of Artificial Intelligence and Robotics for Society (AIRS) and its SciArt Co-lab (archive page).
What it is. A video installation drawn from laboratory research on magnetically controlled micro-nano robot collectives. It uses high-magnification microscope recordings of magnetic microparticles that self-assemble and move together under an oscillating magnetic field. The catalog text says the motion it shows, growth, tilting, travel and swarm coordination, is the direct physical result of the magnetic control system, enlarged to cinematic scale.
The question it asks. What counts as animation when nothing is drawn or simulated? The jury praised it for treating animation as "a process of revealing, not fabricating."
What to look at closely. Order without a character. With no figure to follow, the viewer reads behavior: how the swarm organizes, holds and changes. The archive page links a video of the work.
How the six works compare
| Work | Artist | Approach | Award | Where to see it |
|---|---|---|---|---|
| Deer Hunter | Andrew Herzog | Synthetic documentary | Golden Nica | Festival page, artist site |
| Le Miroir | Camille Sauer | Animation from real processes: performed in real time with physical sets | Honorary Mention | Artist page with teaser |
| Spray and Pray | Veneta Androva | Synthetic documentary | Honorary Mention | Artist page with trailer and screenings |
| Ping Pong | Tianji Yu | Questioning AI images: a painted documentary of a conversation with AI | Honorary Mention | Labocine listing (not publicly available) |
| In Writing "Hurrah for Freedom" | Jungwoo Lee | Questioning AI images | Honorary Mention | Archive page |
| Micro-Nano Landscape | Suxuan Jiang | Animation from real processes | Honorary Mention | Archive page, video |
The approach column is our reading, not the festival's. Several works sit across categories: Deer Hunter and Spray and Pray both question algorithmic images while using them, and Ping Pong is as much about an AI system as it is a documentary.
The same category also gave Awards of Distinction to Esteban Azuela's Aferrado and 2girls1comp's Dancing Plague, and Honorary Mentions to eight more works, all listed on the winners page.
A viewing guide for experimental animation
Use these with any work that mixes recorded, built and generated images.
- Composition. What is in the frame, and what made it: a camera, a scan, a render, a model, a microscope? Where do two of those meet in one image?
- Movement. Who or what is the source of motion: a performer, an animator's keyframes, a physical force, a model's prediction? Does the movement feel authored or found?
- Pacing. Does the edit follow a story, an argument or a process? Where does it slow down, and what does it ask you to look at then?
- Source status. Can you tell archive from home video from generated footage? Does the work tell you, and when? Is the uncertainty the point?
- Sound. Are voices recorded, performed or synthetic? Does the sound confirm the image or work against it?
- Context. Was it made for a cinema, a gallery, an installation or a festival program? What changes if you watch it on a laptop?
- Feeling. What did you feel, and which of the above produced it?
As an exercise, answer the source status question shot by shot for two minutes of any one work.
Watchlist and official links
- Prix Ars Electronica 2026 winners: every New Animation Art award with short jury notes.
- 2026 jury statements: the full New Animation Art statement.
- New Animation Art category description: what the category accepts.
- Press release, June 22, 2026: winners, jury and festival plans.
- Deer Hunter: archive, festival page, Andrew Herzog.
- Le Miroir: archive, Camille Sauer.
- Spray and Pray: archive, Veneta Androva.
- Ping Pong: archive, DocNomads, Labocine.
- In Writing "Hurrah for Freedom": Cinema and History: archive.
- Micro-Nano Landscape: archive, video.