REAS

COMPRESSED CINEMA

As a body of work, Compressed Cinema (2018–2023) reimagines the structures of film through the logic of generative systems. At its center are five videos, created in collaboration with composer Jan St. Werner, and five corresponding series of still images. Together, these works ask what it means to distill the richness of cinema into another register—shortened, compressed, and transformed.

Each video in the series reduces a feature-length film into fewer than ten minutes, not through conventional editing, but by drawing on a field of images synthesized by generative algorithms. What results is not simply a shortened narrative, but a new order of perception, in which each film’s continuity is tested against the nonlinearity of code.

In Compressed Cinema, sound and image are treated as inseparable partners. Werner’s compositions, crafted from electronic timbres and granular transformations, mirror the instability of Reas’ visuals. As the imagery shifts from faces to textures, from landscapes to near-total abstraction, the sound oscillates in tandem to create passages of recognition, rupture, and dreamlike suspension. This interplay is central to the project: cinema is not merely compressed in duration, but reconfigured as an audiovisual field where perception itself is stretched, destabilized, and intensified. In this way, film reveals itself less as a form of storytelling, and more as an experience of vision and sound under pressure.

Untitled Film Stills

Overall, the Compressed Cinema series began not with moving images, but with stills. From each film source, thousands of frames were extracted as PNG image files and used to train a generative model—which then synthesized new images, each of which carries a trace of the original film, without reproducing it. Released as five sets of images corresponding to five videos, the Untitled Film Stills form the foundation of Compressed Cinema.

Together, the stills embody a spectrum of transformation. Some are tethered to representation: a face hovering at the edge of recognition, a doorframe emerging from grain, a body moving through shadow. Others dissolve into dense textures, painterly surfaces, or jagged fragments that retain almost nothing of the original shot. However, most hover in between, as hybrid forms that tease at recognition while simultaneously slipping into abstraction. The stills hold the tension of cinema, while reducing it to sensation—color, mood, and symbolic charge—rather than the continuity of plot. In this sense, they suggest that the essence of cinema is not narrative progression but the persistence of perception itself.

Untitled film still 3.1, 2019/2023.

Untitled film still 3.15, 2019/2023.

Untitled film still 5.1, 2019/2023.

Untitled film still 5.8, 2019/2023.

Crucially, the series of stills was produced independently, before the corresponding videos. Each still image serves as a fixed point within the videos, with new image sequences interpolating through these anchors to generate motion. Reas and his technical lead on the project, Hye Min Cho, wrote custom software to choreograph the movement through each GAN model’s latent space. Therefore, the role of these images is dual: they are autonomous works on their own, as well as the structural framework through which the compressed films unfold. This inversion—where stills define motion rather than documenting it—underscores the project’s rethinking of cinematic order.

Taken together, the Film Stills imagine an alternate history of cinema in which films are remembered not for their storylines, but for the atmospheres they generate. They condense the symbolic and affective weight of the originals into images that are at once familiar and estranged, reconstituting cinema less as narrative spectacle, and more as a mode of visionary image-making.

Untitled Videos

The Untitled series consists of five videos: Untitled 1 (No. Nothing.), Untitled 2 (Kiss me.), Untitled 3 (I withdraw.), Untitled 4 (Two dead!), and Untitled 5 (Not now. No, no.). These are not remakes or adaptations, but rather condensations that compress feature-length films into concentrated audiovisual experiences. While each video pursues its own distinct tempo and atmosphere, together they form a sequence that continually builds on ideas explored in the prior works.

What emerges from these compressions is a continual oscillation. Figures and spaces flicker into view, only to dissolve back into abstraction, while textures coalesce unexpectedly into legible form. This rhythm of recognition and loss is central to the experience of the videos. They remind us that film is always both representational and abstract, consisting entirely of frames of light and shadow that cohere into meaning only through our perception. Werner’s soundscapes sharpen this effect: noise surges when the image collapses, tones stretch when figures begin to form, and rhythms falter as a room melts into texture. Together, image and sound generate a cinematic space that is familiar yet unstable, compressed into dreamlike intensity.

excerpt from untitled 3 (i withdraw.), 2020.

excerpt from untitled 5 (not now. no, no.), 2020.

Technology as Timeline

The five still and video works of Compressed Cinema also trace, indirectly, the rapid evolution of generative image systems. The first two bodies (Film Stills 1–2 and Videos 1–2) were started in 2018 with DCGAN, one of the earliest architectures for generative adversarial networks. The resulting images are low-resolution and filled with visible artifacts—qualities that Reas does not conceal, but instead embraces. This gives the early works a jagged, experimental character that evokes cinema torn apart and crudely reassembled.

Installation view of Untitled film stills 4.14 and 4.16 in conjured terrain at dam projects. berlin, germany. 2023.

Installation view of conjured terrain at vellum. los angeles, ca. 2023.

By 2019, when the later Compressed Cinema works were in production (Film Stills 3–5 and Videos 3–5), Reas’ studio was using the newer Progressive Growing of GANs, a technique capable of producing higher-resolution outputs and smoother transitions. As such, the images in the later videos are more fluid and continuous (though uncanny hybrids and distortions persist). The rapid progression from coarse fragments to more coherent surfaces is inscribed in the project itself, embedding a history of machine learning within the artworks.

Just as early cinema carried the flicker of primitive projectors, Compressed Cinema bears the imprint of its generative instruments. In both cases, the artifact is not simply a limitation, but a declaration of medium: flicker was the condition through which cinema first appeared, and the glitches and seams of GANs mark the conditions of its algorithmic counterpart. The traces of training and resolution are thus not flaws to be erased, but the very texture of the work, and reminders that the history of cinema is always entwined with the history of its machines.

Compression as Concept

The title Compressed Cinema carries multiple resonances. It refers to duration: feature-length films contracted into shorts. But it also signals a different order of perception: cinema reimagined as concentration rather than extension. Compression here is not simply reduction, but intensification: as story and character dissolve, they leave in their place fields of sensation and symbolic resonance. In this sense, the works echo the way films are often remembered—not as continuous narratives but as atmospheres, flashes, and fragments that linger after the story dissolves.

SCreening of COMPRESSED CINEMA videos at babylon. berlin, germany. 2023.

This act of compression connects the works to traditions of experimental film. Ken Jacobs’ Tom, Tom, the Piper’s Son (1969)—which expanded an eight-minute film into more than an hour through rephotography and repetition—is a key precedent. But where Jacobs expanded, Reas contracts. Instead of pulling apart frames to reveal hidden detail, Compressed Cinema distills entire features into concentrated experiences. Both strategies invert mainstream cinema’s seamless flow, foregrounding its constructedness and exposing what lies beneath its narrative surface.

Alchemy

Compressed Cinema is not a demonstration of new image technologies. It is a reflection on how cinema might shift when refracted through generative systems. The works reinforce the idea that cinema is not only about narrative or representation, but also about affect, atmosphere, and the unstable interplay between the two. Paired with Jan St. Werner’s soundscapes, these videos foreground the audiovisual nature of cinematic experience, where image and sound fuse together.

The project also insists on authorship and aesthetic choice. While generative systems open fields of possibility, the works are still shaped by decisions: what data to include, what images to select, how to navigate the latent space, and how to choreograph sound and image together. In this way, the system functions as an instrument, and the art lies in how it is played.

Overall, Compressed Cinema is about revealing cinema’s essence in another form; it is a meditation on memory, perception, and the mutability of media in the age of generative systems.

Related Artworks
Browse the Compressed Cinema category on the REAS INDEX.

Further Exploration
Bazin, André. What Is Cinema? Translated by Hugh Gray. 2 vols. Berkeley: University of California Press, 1967–1971.

Jacobs, Ken. Tom, Tom, the Piper’s Son. 16mm film, 1969.

Reas, Casey. Making Pictures with Generative Adversarial Networks. Montreal: Anteism Books, 2019. Introduction by Nora N. Khan.

Reas, Casey, and Allison Parrish. Compressed Cinema. Montreal: Anteism, 2023.

Youngblood, Gene. Expanded Cinema. New York: Dutton, 1970.

Paul, Christiane. Digital Art. 3rd ed. London: Thames & Hudson, 2015.

Pierson, Michele, David E. James, and Paul Arthur, eds. Optic Antics: The Cinema of Ken Jacobs. New York and London: Oxford University Press, 2011.

Sitney, P. Adams. Visionary Film: The American Avant-Garde, 1943–2000. 3rd ed. New York: Oxford University Press, 2002.

Selected Exhibition History
AI, as Seen at the End of Ownership. Guangdong Times Museum. Guangzhou, China. 21 March – 22 June 2025. Compressed Cinema. [E-25-02]

Topologies of the Real. Meixihu Contemporary Art Museum. Changsha, Hunan, China. 10 November 2024 – 10 March 2025. Compressed Cinema. [E-24-07]

Infinite Games: Hello, World! 798 Art Zone. Beijing, China. 16 November – 25 December 2023. Curated by Baoyang Chen and Emma Cheng. Included Compressed Cinema. [E-23-12]

Conjured Terrain. Vellum. Los Angeles, CA. 6 – 29 October 2023. Two-part exhibition starting in Berlin and continuing in Los Angeles. [S-26]

Conjured Terrain. DAM. Berlin, Germany. 5 – 23 September 2023. Two-part exhibition starting in Berlin and continuing in Los Angeles. Debut of Untitled Film Stills Series 3, 4, 5. [S-25]

Compressed Cinema. Anthology Film Archives. New York, NY. 11 November 2023. U.S. theatrical premiere of the five Compressed Cinema videos. [E-23-11]

Compressed Cinema. Babylon. Berlin, Germany. 10 September 2023. Theatrical premiere of the five Compressed Cinema videos. [E-23-07]

AI Delivered: The Abject. Chronus Art Center (CAC). Shanghai, China. 3 July – 17 October 2021. Compressed Cinema videos. [E-21-04]

Alchemical. bitforms gallery. New York, NY. 8 January – 14 February 2021. Debut of Compressed Cinema videos created with Jan St. Werner. [S-22]

Thin as Thorns, In These Thoughts in Us: An Exhibition of Creative AI and Generative Art. Honor Fraser. Los Angeles, CA. 8 September 2020 – 20 February 2021. Curated by Kenric McDowell and Paul Young. Screening of Compressed Cinema videos. [E-20-04]

Compressed Cinema. Walker Art Center Virtual Cinema. Minneapolis, MN. 10–18 November 2020. Curated by Michael Walsh. Screening of Compressed Cinema videos. [E-20-05]

Compressed Cinema. DAM Projects. Berlin, Germany. 22 March – 4 May 2019. Debut of Untitled Film Stills and early Compressed Cinema videos. [S-21]