Alternative Version: https://f.io/X34iBhjs

Predator is an animated short film about perception, prejudice, and the media staging of fear. At the center of the story is a young swimmer named Reese, who finds himself alone in the open water. When a shark’s dorsal fin appears behind him, he believes he is being chased by a dangerous predator. His fear continues to grow as the shark seems to come closer and closer.

At first, the film deliberately follows the expectations of classic shark and horror films. Through tight framing, fast movements, suspenseful editing, and threatening sound design, the impression of an imminent attack is created. Both Reese and the audience assume that the shark is deliberately hunting the swimmer.

Only later does the meaning of the situation change. The shark dives beneath Reese and is actually pursuing different prey. The danger built up beforehand therefore turns out to be a misinterpretation. Although the swimmer has indeed encountered a shark, he was not its intended target.

The events are then placed in a different context once again through a news report. The reporter Liv covers the incident dramatically and once again presents the shark as a dangerous attacker.

At the end, the camera pulls out of the television screen. Only then does it become visible that the entire news report has been watched by a person in a living room decorated with shark merchandise. At the very end, the viewer switches off the entire image. This opens up another level of perception: the audience not only observes a story, but also sees how this story is staged, distributed, and consumed.

The title Predator therefore does not refer exclusively to the shark. Through five different shifts in perspective, it instead raises the question of who or what is actually perceived as the “predator” within the story: the shark, the swimmer’s fear, the sensationalist reporting, or the audience’s media-shaped perception.

Starting Point and Idea

The starting point of the project was the question of how strongly the meaning of an event can change through different perspectives. Depending on the point of view, the available knowledge, and the way it is presented by the media, the same situation can be perceived in completely different ways.

The project was created as part of the assignment “Who Is Looking?”. The focus was on examining how perspective, framing, editing, sound, sequence, and context influence the perception of an unchanged event. The aim was not to document reality as objectively as possible, but to deliberately shape it through different ways of seeing.

The so-called Rashomon effect, named after Akira Kurosawa’s film Rashomon, served as a loose inspiration. In the film, the same event is told from several, partly contradictory perspectives. There is no clearly objective version of the truth. Instead, it becomes apparent that observers notice certain details, overlook others, and interpret what they experienced in different ways.

For our film, we therefore wanted to begin by creating a situation that appeared to be clearly dangerous. A swimmer in open water and the fin of a shark immediately create certain expectations. Because of familiar images from films, news reports, and popular culture, the shark is often automatically perceived as the attacker. We wanted to deliberately use this expectation and then gradually question it. The event itself does not change. What changes are the point of view, the available information, and the way the situation is presented.

Content-Related Objective

In addition to its surprising twist, the film addresses a thematic and social question. Predator explores how sharks are portrayed in films, news reports, and popular narratives.

Sharks are often presented as deliberately acting, aggressive predators that specifically hunt humans. However, actual encounters between humans and sharks are rare. Many incidents do not occur because of malice or the deliberate hunting of humans, but may be connected to curiosity, limited visibility, or confusion with their usual prey.

Our film does not address these issues in the form of a documentary reconstruction, but translates them into a deliberately exaggerated fictional story. The shark is not intended to be made harmless or completely separated from its role as a predator. Instead, the film shows that a predator is not automatically an evil attacker that deliberately targets humans.

At the same time, the film questions the role of the media in creating and reinforcing such images. Through dramatic language and sensationalist presentation, a brief, misunderstood moment is turned into a much larger story. Fear is therefore not only communicated, but transformed into a spectacle.

Research and References

Before we began working on the storyboard and visual development, we looked into shark behavior and documented incidents in Hawaii. We were particularly interested in the increased number of recorded shark encounters during the autumn months. This period is sometimes also referred to as “Sharktober.”

During this time, several factors come together. Higher water temperatures, changing movement patterns of the animals, and conditions in coastal regions can increase the likelihood of encounters between humans and sharks.

We also researched tiger sharks, which are regularly found in the waters around Hawaii. Their broad body shape, distinctive striped pattern, and size create a strong visual impact. For this reason, we decided to use a tiger shark as the central animal in the film.

We also researched references for a realistic portrayal of the shark.

Storyboard

After completing the research, we developed an initial storyboard. In it, we defined the basic sequence of events, the shot sizes, and the transitions between the five perspectives.

The storyboard was particularly important because the story strongly depends on which information is revealed at which point. At the beginning, the shark could only be shown clearly enough to confirm its presence, while its actual intention still had to remain unclear. At the same time, the later reveal needed to make it understandable how the misunderstanding had occurred.

During the production process, the storyboard was revised several times. This was not only due to narrative decisions, but also to the technical possibilities and limitations of the AI models used. Some of the originally planned perspectives or movements could not be implemented convincingly, while other shots emerged from unexpected results during the generation process.

Based on the first generated keyframes, we therefore developed a second, more detailed storyboard. It included a partly different sequence and additional intermediate steps. The generated images were not used exclusively as final frames, but also as visual tests that helped us evaluate camera positions and transitions.

As a result, the production process was not linear. The storyboard, generation, and animation continuously influenced one another. When a planned movement did not work, we had to adjust either the prompt, the composition, or the entire shot.

Visual Style

The visual style of Predator combines simplified, graphic character design with comic-like textures and halftone patterns. The images move between illustration, animation, and a generated visual aesthetic.

One of the main challenges was to create a consistent look despite using different production tools. The AI models sometimes interpreted colors, proportions, and textures differently. For this reason, we defined fixed character features and recurring design elements at an early stage.

These included:

a limited color palette, clear silhouettes, simplified facial features, visible outlines, graphic water surfaces, halftone textures, and strong contrasts between the characters and their surroundings.

The water surface was not depicted photorealistically. Instead, we worked with stylized light reflections, simplified wave shapes, graphic shadows, and textures that partly resemble print. This made it easier to combine the different AI-generated and hand-drawn elements.

The color design also supports the different perspectives. The swimming scenes are dominated by intense shades of blue and turquoise. Pink and purple accents highlight Reese. The news section appears more controlled and staged, while the living room at the end has its own slightly exaggerated shark-fan aesthetic.

Character Development

Reese – The Swimmer

Reese is the central character through whom the audience identifies with the story. His fear and limited perception shape the first part of the film.

In terms of design, Reese was intended to appear young, athletic, and generally likeable. At the same time, his silhouette had to remain clearly recognizable during fast swimming movements and in wider shots. His striking pink hair creates a strong contrast with the blue water and helps the audience visually follow him within the moving environment.

The character sheet includes different views, body postures, facial expressions, color combinations, and details of his swimwear. Various expressions ranging from uncertainty and tension to panic were particularly important.

During production, the character had to be adjusted several times. Since Reese is shown shirtless, relatively young, and in a dangerous situation, several generations were rejected or restricted by the AI platforms used. We therefore modified his appearance and made sure to clearly describe him as an adult swimmer in the prompts.

These technical limitations also influenced the final visual design of the character.

Liv – The Reporter

Liv appears in the news segment of the film. She does not present the event to the audience neutrally, but as part of a dramatized news report.

Her design follows the same graphic style as Reese, but differs through a clearer and more controlled body language. While Reese reacts frantically and emotionally, Liv stands firmly in front of the camera and speaks with professional confidence.

Her character sheet includes different views, facial expressions, a color palette, and the clothing used for the news report. The design was intended to immediately communicate that she is a reporter without making her appear too realistic or visually disconnected from the rest of the film.

Her animation focused less on complex body movement and more on believable speaking movements, small gestures, and a stable position in front of the camera.

The Tiger Shark

To develop the shark, we first created a simple 3D model. It helped us better understand the body shape, proportions, and spatial appearance of a tiger shark.

The model was not used as the final character in the film, but as a visual reference for later drawings and AI-generated images. The broad shape of the head, the position of the dorsal fin, the movement of the body, and the characteristic striped pattern were particularly important.

Based on this, we developed a style guide for the shark. In it, we defined the color palette, texture, halftone pattern, and level of abstraction. The shark needed to appear threatening enough to make the initial misinterpretation believable. At the same time, it could not look like a classic, exaggerated movie monster.

The stripes and halftone pattern became important graphic elements. They connect the shark to the overall illustrative style and prevent it from appearing too photorealistic through AI generation.

Production Process

Development of the Keyframes with ChatGPT

At the beginning of the image production process, we uploaded our storyboard and character sheets to ChatGPT. Based on these, we developed the required keyframes step by step, which can already be seen in our second storyboard.

For each scene, the composition, camera position, viewing direction, character pose, lighting mood, and important continuity features were defined. Since the film consists of many short shots, it was important not to generate each scene independently. Instead, we worked with existing images as visual references and developed the following frames from them.

This approach helped us maintain the most consistent style possible across different images. Nevertheless, deviations kept occurring.

Some keyframes could therefore be used directly. Others had to be revised, expanded, or completely redrawn in Procreate or Photoshop.

Initial Animation Tests with Kling

For the first animation tests, we used Kling. To keep the movement as controlled as possible, we initially worked with partly identical start and end frames for the swimmer. This was intended to prevent the character or the composition from changing too much during the animation.

For the shark, we sometimes used different start and end frames. This allowed us to define a clear direction of movement and, for example, determine whether the shark was swimming towards the camera, remaining beneath the water’s surface, or falling back into the water.

Kling produced interesting results for some camera and water movements. However, clear problems occurred with complex body movements. Creating a natural swimming motion was particularly difficult.

The Swimming Motion as a Central Challenge

Animating Reese quickly proved to be one of the greatest technical challenges. In many attempts, it did not look as though he was actually swimming. Instead, he sometimes appeared to be crawling or walking through the water, or being moved by an invisible force.

His arms, hands, and body posture also changed during the movement. At times, Reese suddenly looked directly into the camera, even though his viewing direction had been clearly defined.

We therefore tested different wordings, movement descriptions, camera perspectives, and reference images. Terms such as “fast freestyle swimming,” “strong forward propulsion,” “horizontal body position,” and “natural swimming cycle” were partly helpful, but did not fully solve the problem.

This difficulty showed that AI animation is not only a matter of finding the right prompt. Some movements require a spatial and physical understanding that the models we used were only able to reproduce to a limited extent.

Switching to Seedance 2.0

Compared to the previous tests, Seedance 2.0 was able to create the swimming motion more convincingly. We therefore switched tools during the production process.

The shark’s movements could also be controlled more effectively in some cases. More stable results were achieved particularly with fast forward movements, underwater shots, and transitions between the water’s surface and the underwater perspective.

However, the switch did not solve every problem. Seedance 2.0 also produced unusable variations.

Nevertheless, the large number of failed attempts was part of the learning process. By comparing the models, we were able to better assess which tool was best suited for each type of shot.

Manual Animation in Procreate Dreams

Some scenes could not be implemented convincingly with either Kling or Seedance 2.0. We therefore drew and animated these sequences frame by frame in Procreate Dreams.

Manual animation gave us significantly greater control over timing, silhouette, and direction of movement. At the same time, it was more time-consuming. It was therefore used specifically for shots in which the AI-generated results were particularly unstable or a very precise movement was required.

The hand-drawn sequences then had to be visually adapted to the generated clips. To achieve this, we used the same color palette, similar outlines, and the halftone pattern.

The combination of AI-generated and hand-drawn scenes was therefore not treated as a workaround, but as a hybrid production process. The most suitable tool was selected depending on the requirements of each scene.

Keying and Connecting Different Clips

Another challenge was connecting independently generated clips. Even when the beginning and end matched in terms of content, the water movement, positions, or lighting conditions often differed.

To combine two sequences more smoothly, we drew an additional keying or transition element by hand. This allowed us to blend individual areas of the image in a controlled way and coordinate the movements more effectively.

This step shows that a functioning AI-generated clip does not automatically create a functioning film scene. A coherent sequence only emerged through the combination of transitions, editing, compositing, and manual corrections.

Editing, Compositing, and Sound Design

The final clips were edited, adjusted, and combined into a coherent dramatic sequence in Premiere Pro. We mainly used the most stable sections of the AI-generated animations and corrected the framing, speed, colors, and transitions.

The sound design was created using Logic and Maschine Studio and supports the shifts in perspective. Water sounds and suspenseful music initially intensify the sense of danger, while the sound takes on a new meaning during the reveal and the news report. The final transition into the living room makes it both audible and visible that the events are being communicated through the media.

Manual Correction of AI Errors

In the final stage of production, we manually corrected smaller AI-generated errors. These included unwanted objects, unstable shapes, and elements that changed between individual frames.

One example was a group of seagulls in the upper-right area of the image. They were displayed incorrectly during the movement and distracted from the actual scene. We therefore removed them manually from the affected frames.

Challenges and Discarded Approaches

One of the greatest challenges was maintaining visual consistency. Even though we worked with character sheets and reference images, the models repeatedly changed faces, body proportions, clothing, and colors.

The swimming motion was particularly problematic. Many attempts looked more like crawling or walking underwater. The shark also sometimes moved in anatomically illogical ways or lost its characteristic appearance during the animation.

There were also error messages and restrictions related to the depiction of Reese. The combination of his youthful appearance, bare upper body, and dangerous situation caused several tools to block the generations. We therefore had to present the character more clearly as an adult and visually adjust individual scenes.

Another discarded approach was to create certain movements entirely within a single long clip. This usually resulted in too many changes and artifacts. It proved more successful to divide the movement into several short clips and connect them during editing.

Not every generated variation could be used. In some cases, additional animals, incorrect viewing directions, deformed bodies, or objects that were not present in the original image appeared. Nevertheless, these failed attempts were helpful because they showed us which information the AI misinterpreted and which scenes required a manual approach.

Earlier Exercises and Experiments from the Semester

The development of Predator was prepared through several smaller exercises and experiments during the semester. These assignments helped us test different perspectives, visual moods, and AI-assisted production methods.

Day in a Life

In the “Day in a Life” exercise, we developed short sequences from the perspective of a fish and a cat. We worked with viewing directions, spatial transitions, and different camera positions.

The shots included a view from inside an aquarium, a shot of the water’s surface from below, and a cat’s eyes in front of the glass. For the cat, we created scenes of it waking up, standing in a doorway, and observing the aquarium.

This exercise was particularly relevant to *Predator* because it was the first time we explored a non-human perspective in greater depth. The question of how an animal perceives its surroundings and how strongly a scene changes through a shift in perspective later became a central element of the final film.

The depiction of water and the transition between an inside and outside perspective also provided initial visual experience for the later underwater scenes.

Light & Perspective

In the “Light & Perspective” exercise, we developed two scenes with two versions each. Through lighting, color, camera position, and editing, the situations took on a different meaning in each version.

One version appeared warm, calm, and ordinary. The other appeared cold, distant, and observational. Although the action remained almost identical, the emotional effect changed significantly.

This insight directly influenced *Predator*. In the film, the same situation is also interpreted differently depending on the framing, sound, and context.

Film-Opening- und Storyboard-Experiment

In another assignment, we developed a storyboard for a film opening in which a character runs through a mine. A drop of water transforms the main character into a fish-like figure before the situation changes through a gunshot, a fall, and the transition into an underground underwater world.

The focus of this exercise was on storyboarding, building suspense, and creating visual transitions.

For Predator, working with keyframes and transitional images was particularly helpful.

Reflection

What changed between the initial idea and the final result?

The story remained the same, but the production process became much more hybrid. Many scenes had to be manually corrected, redrawn, or animated frame by frame. The storyboard was also adapted several times to the possibilities of the tools used.

Which perspective changed the meaning the most?

The strongest shift occurs in the moment when it becomes clear that the shark is not chasing Reese. The later media perspective changes the event once again by simplifying and dramatizing it.

Which technical process changed the project the most?

The shift from a purely AI-based production to a combination of AI, manual animation, editing, and sound design was crucial. Procreate Dreams and Premiere Pro were particularly helpful in controlling unstable results.

What worked particularly well?

The clear structure of the five perspectives, the character sheets, and the consistent visual style worked particularly well. The targeted use of different programs for different tasks also proved effective.

What was difficult or did not work?

Natural swimming movements, stable anatomy, and character continuity were particularly problematic. Many attempts were unusable, resulting in a high investment of time and money.

What did we learn during the project?

We learned that AI film production involves much more than generating images. Storytelling, animation, editing, sound design, and manual post-production remain essential. At the same time, errors and unexpected results can open up new approaches and solutions.

What could be helpful for others?

Clear character sheets, color palettes, and reference images are helpful. Complex movements should be divided into short sections, and AI-generated clips should be treated as raw material from the beginning. Failed attempts should also be documented, as they provide important insights into the process.

Conclusion

Predator shows how strongly fear can be created through perspective, missing information, and media staging. The film begins with the familiar image of the dangerous shark and then gradually challenges and deconstructs this image.

At the same time, the project was an intensive exploration of the possibilities and limitations of AI-assisted film production. We learned that a convincing AI film is not created through generation alone. It requires a clear story, visual planning, design decisions, numerous tests, and extensive manual post-production.

The production process was longer, more complex, and more expensive than initially expected. Many clips had to be generated several times, adjusted, or completely replaced. At the same time, working with different tools opened up new possibilities for images and transitions that we might have approached differently using a purely traditional production process.

One of the most important realizations was that human decisions remain central throughout the entire process. The AI generated different variations, but we had to select, evaluate, correct, and decide which images actually told the story.

Despite the technical difficulties and numerous failed attempts, we are very satisfied with the final result. The film combines a critical examination of the portrayal of sharks with an experimental production process and, at the same time, makes visible the many different steps involved in creating an AI-animated short film.