Workshop at Ars Electronica 2026 : Fluidity of Images

Surrealist Juxtaposition and Image Mixing in Generative AI

Projektbetreuung

Richtung

Künstliche Intelligenz/AI
Zeichnung

Projekt Art

Fluidity of Images: Surrealist Juxtaposition and Image Mixing in Generative AI

The workshop at Ars Electronica 2026 in the framework of "HAMLET AI is the Thing" activities is conducted by Mert Akbal with the assistance of Michael Schmitz.

What happens when a machine mixes paintings without text prompts? Fluidity of Images explores the juxtaposition space between images: the territory between two source works where they combine to different extents to reveal a spectrum of new forms, condensations and compositions. It situates this practice within a broader continuum of procedural and collective artistic practices, where agency is distributed across processes, tools and collaborators.

Workshop Overview

Workshop Aims

To demystify AI image generation through surrealism, by showing the parallels between surrealist methods and generative workflows
To demystify visual art through surrealism by demonstrating that everybody is able to communicate through drawings when approached playfully
To think about the future of work collectively and collect opinions and ideas for the HAMLET project

In the workshop two surrealist drawing games were played by participants. Between the exercises similarities between these games and the working principles of generative AI models are introduced, using open source models and interfaces. Finally, short interviews are conducted with participants about their views on the future of work, especially in the context of the growing role of AI.

Workshop Activities

Exercise 1: Frottage and Pareidolia

The first surrealist game was frottage. Participants are asked to rub charcoal on paper with a piece of plant leaf placed underneath, revealing plant structures. Meanwhile they are prompted to think about what they want to work on in three years.

The frottage game is based on the concept of pareidolia, the perception phenomenon of seeing faces and bodies of humans and animals in seemingly random formations like rocks, clouds and wood. Participants are asked to see an image of their imagined future work in the drawing and mark it with another color. This exercise aims to reveal simple ideas on the future of work while working on something as tactile as rubbing charcoal on paper. Answers are then collected through interviews and transcribed.

In the next step, the analogy between frottage and denoising in generative AI is demonstrated. The concept of denoising in image generation is explained based on the pareidolia phenomenon and the hands-on frottage exercise. An image-to-image workflow is demonstrated on the ComfyUI interface with a Stable Diffusion 1.5 model. Three parameters of this workflow are explored: number of generation steps, denoising strength, and different word prompts.

Exercise 2: Collective Drawing and Image Mixing

The second surrealist game was collective drawing. Each participant received a sheet of paper and pencils and crayons for drawing. The collective drawing process resembles production on an assembly line in a factory. Each time, participants forwarded the paper in their hand with instructions to their neighbors, along with a written instruction on what to add to the drawing in their turn. After completing their drawing iteration by adding or changing elements, they photographed the current status with their phones and forwarded it with a new instruction to the next person.

In this workshop exercise the discussion is framed around the future of work. Participants are directed in the first four iterations by the workshop moderator to draw a horizon, draw a vertical environment, add a human figure (themselves), and add what this person is working on. Starting with the fifth iteration, they are asked to negotiate the working conditions of this person by giving instructions to their neighbors as they forward the drawing.

After finishing the drawings, they are loaded as image inputs into the Flux.Redux image mixing workflow in the ComfyUI interface. This demonstrates how elements from multiple images can be mixed to generate different meanings each time.