Post-Processing Analysis pt.2 - Setting up the Project

Research Summary

This research explores post-processing in stylized 3D horror video games and which post-processing methods are most effective in creating the emotion of 'unease'.

In part 1 of this analysis I evaluated the post-processing elements and techniques used in six stylized 3D horror video games. Based on the results, I chose three unique post-processing styles to create within Unreal Engine 5: A pixel shader (emphasis on grunge and grainy visuals), a toon shader (emphasis on exaggerated outlines and sharp edges) and a painterly shader (emphasis on the hand-drawn look, crosshatched shading). The purpose of these unique styles is to compare which are most effective in creating the emotion of unease, and why. This post discusses and justifies the decisions made in setting up the project scene.

Choosing the Scene

The function of the scene is to emulate a video game environment to overlay my post-processing profiles onto. I chose to source the models and textures for the scene from 3D asset sharing websites (Fab, Itch.io) in order to get straight to the research and focus on developing the post-processing methods.

There were a few elements to consider when choosing the base models and scene. We can assume stylized post-processing would be used alongside stylized models. However, to avoid the models affecting the results of my study I decided to look for models that were stylized but not to the extreme. I avoided photo-realistic models and looked out for ones with a 'basic' appearance - not so exaggerated that they would distract future surveyees from the post-processing's effect on the scene. Additionally I did not want the models to be inherently horror themed in case they overpowered the post-processing in contributing to the feeling of 'eeriness'.

I looked for asset packs instead of compiling individual assets from different sources to ensure the scene was consistent in art style - this better emulates how a real game scene would look. I considered using a pre-made scene, but modular asset packs provided more flexibility and control over the scene layout.

In summary, the scene that would best support my research needed to be:

• Generic in theme

• Minimal exaggeration

• Not photo-realistic

• Fit all 3 post-processing methods

• Work with Unreal Engine

• Permission to use from original creator

It proved difficult to find a free asset pack that did not have an apparent theme, so I compared the pros and cons of three asset packs.

A B C

(Gigel3d, 2025)

• ↑ Stylized

• ↓ Slightly thematic (medieval)

(EmacEArt, 2026)

• ↑ Stylized

• ↓ Too thematic (desert, western)

(Loafbrr. n.d.)

• ↑ Neutral theme

• ↓ Less stylized

	A	B	C
Generic in theme	✓	X	✓
Minimal exaggeration	✓	✓	✓
Not photo-realistic	✓	✓	✓
Fit all 3 post-processing methods	✓	✓	✓
Work with Unreal Engine	✓	✓	✓
Permission to use from original creator	✓	✓	✓

I chose to omit asset pack B as asset pack A and C better fit my criteria. Although, I speculate that choosing an art style with minimal exaggeration is more important than choosing one that is generic in theme, because the horror genre can be applied across any theme or setting. For example, the horror games analyzed in the first blog are all different settings, but all successfully create an 'eerie' gameplay experience. Regardless, I still decided to exclude asset pack B just to be safe. Based on these results, I considered a couple different options.

Option 1 is to set up two scenes - one using asset pack A and the other using asset pack C. Initially I wanted to keep the scene consistent to avoid elements such as level design and theme affecting the research outcome. In order to still achieve this, the three post-processing styles will be compared across asset pack A only. Then in a separate part of the survey all three post-processing styles will be compared across asset pack B only. This would also give insight into whether the scene has affected the 'eeriness' by comparing the survey answers across the same post-processing styles over the two scenes.

Option 2 is to only set up one scene using asset pack A or C. This is more in-line with my original plan to use one consistent scene. This also means more time dedicated to developing the post-processing styles.

I initially decided to set-up both scenes. However, I ran into some issues with setting up asset pack C - mainly due to the fact that it is an interior scene rather than exterior. Since it is interior, I cannot rely on the 'sky light' to illuminate the scene. I would need to set up interior lighting which would make the process of setting up the scene take longer. Another issue was with collisions. I would need to individually adjust the collision boxes of each model in the asset pack to enable the player to move around the interior. This is doable but time-consuming, therefore I chose to use only asset pack A for now.

Project Set-up

I used Unreal Engine 5's 'First Person' template as a base for the project. This template provides a pre-made player and camera controlled with keyboard and mouse inputs. I chose first-person perspective over third-person as the focus is on how post-processing affects the environment's mood. Having a visible character was unnecessary for the purpose of my research and could become a distraction from the environment.

The only adjustments I made to the player were lowering the walk speed and removing the visible player arms. The default walk speed felt too fast for the player to fully focus on their surroundings. Additionally, horror games often feature a lower walk speed, although to avoid this affecting the outcome of my research (players finding the level more eerie due to a slower walk speed), the speed is kept consistent across all post-process styles.

The player arms were removed to avoid distraction from the scene, although removing the player arms may result in a more 'spectator-like' and therefore less 'real' and immersive player experience (potentially making the level less eerie due to the feeling of not inhabiting a human body). I could test the difference between how arms versus no arms affect the player's experience, but in order to keep this research focused solely on post-processing I decided against this.

Scene 1

This scene is made from the assets provided in asset pack A (Gigel3d, 2025). There was no specific structure or rules I followed when arranging the scene. The aim was to keep the scene basic to avoid over-powering the post-processing effects. There are no interactable props and the houses cannot be entered. The only control the player has is walking and jumping around the level. To stay on topic, this blog will not delve into the technical steps taken to set-up the level layout in further detail.

Conclusion

Overall, I believe the models in asset pack A fit well in terms of visual style due to being non-photorealistic while also not too abstractly or exaggeratedly stylized, despite having a 'medieval' theme. Asset pack C may have been a better choice, which I would have chosen if not for the time-consuming set-up. Although, I suspect the 'medieval village' theme of asset pack A will not have a large effect on the outcome of this research.

The scene on its own intentionally does not create an 'eerie' mood, as the focus for this research is on achieving that mood through post-processing effects. The next step will be creating the 3 post-processing styles mentioned earlier. The following blog posts will focus on the development and choices made in designing these post-processing styles.

References

EmacEArt. (2026). Free Low Poly - Raft on the desert. Fab. https://www.fab.com/listings/725e2f8e-18dc-432d-b427-86431d1fa9f2

Gigel3d. (2025). RPG Poly Pack - Lite. Fab. https://www.fab.com/listings/ea7f5167-a797-45fe-96d4-efb9f3aecc20

Loafbrr. (n.d.). Interiors A. Itchio. https://loafbrr.itch.io/interiors-a