5.1 Arnold Rendering & Optimisation
Establishing the Rendering Workflow
After completing the main animation and lighting setups, I began preparing the scenes for rendering using Arnold in Autodesk Maya.
My main objective was to achieve consistent image quality across the different shots while keeping rendering times manageable.
This was particularly important because the film combines two environments with different lighting conditions, materials and levels of complexity.
I wanted to maintain enough visual detail for the characters and environments to appear convincing without introducing unnecessarily long rendering times.
Render Settings and Sampling
During the initial rendering tests, I experimented with Arnold’s sampling settings to balance image quality and rendering performance.
I initially used a Camera AA value of 5 before increasing it to 8 for the final renders, as some of the earlier results contained visible noise, particularly in areas involving shadows and indirect lighting.
To further reduce the grainy appearance, I used Arnold’s OptiX Imager Denoiser, which helped smooth out noise in the rendered images.
Increasing the sampling value improved image quality but also increased rendering times. I therefore compared test renders to evaluate whether the visual improvements justified the additional processing time.
Combining higher sampling with denoising allowed me to work towards cleaner images while developing a better understanding of the trade-off between rendering quality and efficiency.

Rendering Image Sequences
I rendered the animation as individual PNG image sequences.
Some sequences required a large number of frames, making the rendering process particularly time-consuming.
One of the most frustrating aspects was discovering mistakes after completing a render, as this often meant returning to Maya, correcting the issue and rendering the entire shot again.
For example, in one sequence, a car did not move far enough within the shot, while in another, part of a background character’s body became unintentionally visible due to an issue with the clothing mesh.
Although working with PNG sequences allowed me to inspect the rendered output before importing it into Premiere Pro, it did not prevent the need to repeat lengthy renders.
This experience highlighted the importance of checking animation, camera composition and character geometry more thoroughly before committing to a full render.
5.2 Materials & Texture Refinement
Working with Imported Materials
Throughout production, I encountered several inconsistencies in the materials and textures of imported 3D assets.
Because the environments and characters originated from different sources, their shaders did not always respond consistently to the lighting setups.
Some surfaces appeared excessively reflective, while others displayed incorrect colours or required adjustments to their transparency.
I used Maya’s Hypershade and Attribute Editor to investigate the relevant material properties and texture connections.
Rather than modifying the entire visual style of the scene, I focused on correcting individual assets that appeared noticeably inconsistent with their surroundings.
Adjusting Excessively Glossy Materials
One of the most noticeable issues involved the background characters and certain objects in the city environment.
Several imported materials appeared unusually glossy, creating strong reflections and highlights that made some surfaces appear almost plastic-like.
After investigating the shader settings in Maya, I identified that the Cosine Power value was too high on some of the affected materials.
This setting influenced the concentration of the specular highlights, making the reflections appear overly sharp and pronounced.
By reducing the Cosine Power value, I was able to soften the highlights and achieve a more appropriate appearance for the characters and surrounding objects.
I encountered similar issues with environmental assets, including streetlights, which required individual material adjustments.


Refining Traffic Light Materials
To improve the realism of the urban environment, I wanted the traffic lights to appear illuminated rather than simply displaying coloured surfaces.
The original imported materials did not produce a convincing illuminated appearance, and some of the traffic light surfaces appeared almost black when rendered.
I investigated the material settings in Maya and experimented with emission to make the coloured sections more visible.
For example, I increased the emission value of the green traffic light material to approximately 5, which made it appear brighter and more recognisable within the scene.
I applied similar adjustments to the other traffic light materials to achieve a more consistent appearance.
Although this was a relatively small modification, it contributed to making the city environment feel more believable and visually detailed.

5.3 Compositing in After Effects
Exploring Dust and Atmospheric Effects
After animating the building collapse in Maya, I wanted to enhance the sequence by introducing a large dust cloud that would spread across the environment.
My original vision involved a dense, volumetric explosion of dust that would gradually obscure the buildings and create a greater sense of destruction.
I initially experimented with creating the effect myself in Adobe After Effects, exploring how I could simulate the appearance and movement of dust within the composition.
Although I was able to experiment with different approaches, the results did not achieve the density, depth or visual impact I had envisioned.
This led me to investigate alternative techniques, including Unreal Engine’s Niagara particle system, before eventually returning to After Effects with a different compositing approach.

Experimenting with Unreal Engine Niagara
Before developing the final dust effect in After Effects, I explored Unreal Engine’s Niagara particle system as a possible approach to creating a more volumetric dust explosion.
I experimented with a beta version of a dust effect, hoping it would provide more control over the scale and movement of the particles.
However, I encountered several technical difficulties. Initially, the effect was not visible in the scene, which made it difficult to determine whether the problem was related to the effect settings or its placement.
After further experimentation, I also experienced issues with the effect becoming unresponsive, preventing me from adjusting and developing it as intended.
Although I attempted to troubleshoot these problems, I was unable to achieve a reliable result that could be integrated into my building collapse sequence.
I therefore decided to discontinue this approach and focus on compositing existing dust footage in After Effects.
While the Unreal Engine experiment did not produce a usable final effect, it demonstrated the importance of testing unfamiliar tools before relying on them within a production workflow.

Exploring Compositing in After Effects
Following the Unreal Engine experiments, I continued developing the effect in Adobe After Effects.
I explored existing dust footage as an alternative to generating a complete particle simulation.
One advantage of this approach was that I could work directly with the rendered image sequences from Maya, allowing me to evaluate the dust in relation to the actual composition.
However, finding suitable footage was not straightforward.
Some effects appeared too transparent, while others had movement or proportions that did not correspond with the collapsing buildings.
I needed footage that could create the impression of a large cloud expanding across the environment while remaining visually connected to the original scene.
After testing different options, I selected footage that I could further manipulate through compositing.


Developing the Final Composite
Rather than using a single dust layer, I duplicated the selected footage to build up a larger effect.
I experimented with the placement, scale and timing of the different layers to make the dust occupy more of the composition.
One of the techniques I used was After Effects’ Corner Pin effect, which allowed me to distort the footage by adjusting its four corners.
This gave me greater flexibility when adapting the shape and perspective of the dust to the buildings and camera composition.
I also experimented with opacity to control how the layers appeared over the rendered animation.
By combining duplicated footage with these adjustments, I attempted to create the impression of dust spreading from the collapse rather than simply covering the entire frame with a single flat effect.
Blending and Layer Integration
After importing dust footage into After Effects, I experimented with blending modes and layer settings to integrate it with the rendered animation.
Because the footage existed as a separate layer, I needed to consider how its background, colour and transparency interacted with the underlying scene.
I wanted to avoid an effect that looked obviously placed on top of the animation.
I therefore tested different blending approaches and examined how the dust appeared against the buildings and surrounding environment.
This helped me identify the limitations of the footage and determine which adjustments could improve its integration.
Adjusting Scale, Density and Movement
Another challenge was creating the impression of a sufficiently large dust cloud.
Some of the footage I tested appeared too transparent, particularly when placed over the exterior environment.
I experimented with increasing its scale and adjusting its visibility to create a heavier appearance.
However, enlarging the footage sometimes made the effect look flatter and less connected to the perspective of the scene.
I also explored directional movement because I wanted the dust to feel as though it originated from the collapsing buildings rather than expanding uniformly.
These experiments highlighted the difficulty of representing a three-dimensional atmospheric event through two-dimensional footage.


Experimenting with Timing and Transitions
As I continued developing the effect, I also explored how the dust could contribute to the transition between shots.
I experimented with reversing and extending the footage to change the direction and timing of its movement.
I also adjusted opacity to create a gradual fade rather than allowing the dust to disappear abruptly.
This was particularly important because the effect needed to remain connected to the timing of the building collapse.
I considered how the dust could temporarily obscure the environment before gradually disappearing or transitioning into the following sequence.
Although the experiments did not provide the same control as a complete 3D simulation, they allowed me to explore how compositing and editing could work together to create a more continuous visual sequence.
Reflection
The dust experiments became an important example of balancing visual ambition with technical feasibility.
I initially wanted a dense, three-dimensional effect, but creating and integrating this proved more complicated than expected.
Comparing After Effects, Niagara and existing VFX footage helped me evaluate the advantages and limitations of different approaches.
The main difficulty was not simply creating visible dust, but making the effect appear connected to the camera perspective, surrounding environment and timing of the collapse.
This process gave me a better understanding of how environmental effects need to be considered within the complete composition rather than developed in isolation.
5.4 Editing & Sound Design
Assembling the Film in Premiere Pro
After rendering the animated sequences, I began assembling the film in Adobe Premiere Pro.
I imported the PNG image sequences from Maya and arranged the different shots on the timeline.
One of the first technical issues I encountered involved the frame rate.
Although the animation was created at 24 frames per second, Premiere Pro initially interpreted some imported image sequences at 25 fps.
This created a potential inconsistency between the original animation timing and the editing timeline.
I therefore checked the footage interpretation and sequence settings to maintain a consistent frame rate of 24 fps.
This was important because even a small difference in playback speed could influence the timing of character movements and camera animation.
Refining the Pacing
One of the biggest challenges during editing was evaluating how the individual shots worked together.
While certain animations appeared effective when viewed separately in Maya, their timing sometimes felt different once they were assembled into a continuous sequence.
For example, some shots moved too quickly for the audience to observe both the protagonist and the surrounding events.
This was particularly important because the protagonist’s restrained reactions require time to become noticeable.
I therefore experimented with extending several sequences and adjusting the overall rhythm of the film.
In some cases, these changes required returning to Maya to extend camera movements or character animations before rendering additional frames.
Although this added work to the production process, it allowed me to make decisions based on how the film functioned as a complete narrative rather than judging each shot independently.
The editing process therefore became an important part of refining the animation itself.
Music and Sound Design
Since The Observer contains no spoken dialogue, I wanted the soundtrack to contribute to the atmosphere and emotional progression of the film.
I used Stable Audio to generate an original soundtrack based on a detailed prompt describing the intended mood, instrumentation and narrative development.
My prompt focused on a melancholic electronic atmosphere with repetitive rhythms, atmospheric synthesizers and a gradual increase in intensity as the film progresses. I wanted the music to feel immersive and slightly unsettling, reflecting the protagonist’s emotional detachment while contrasting with the increasingly chaotic environment.
The repetitive structure was particularly important because it connected with the concept of routine and normalisation explored during my research.
I was pleased with the generated result. When listening to it alongside the animation, I felt that the soundtrack made the film more immersive and strengthened its emotional atmosphere.
However, I also felt that something was missing. Although the music established the overall mood, it did not fully communicate the physical environment or the events taking place within it.
To address this, I added police sirens, traffic ambience and an explosion sound effect sourced from Pixabay.
These sounds helped establish the urban environment and gave the more dramatic moments a stronger sense of presence.
The combination of the soundtrack and environmental effects allowed me to create a more layered soundscape, where the music supported the film’s psychological atmosphere while the sound effects reinforced the events occurring on screen.
This process demonstrated how sound design could contribute to the narrative beyond simply accompanying the visual animation.
Reflection
Editing the rendered sequences helped me identify pacing problems that were difficult to recognise while working on individual shots in Maya.
I realised that a technically successful animation does not automatically produce an effective sequence when placed alongside other shots.
The editing process required me to reconsider the duration of certain movements and how the audience’s attention shifted throughout the narrative.
It also encouraged me to treat sound as part of the film’s storytelling rather than simply as an additional element introduced after the visual work.
5.5 Final Rendering Challenges & Reflection
Managing Software Instability
One of the most significant challenges during the later stages of production was the instability of Autodesk Maya.
As the project became more complex, I experienced unexpected crashes and slower performance, particularly when working with detailed environments and demanding scene configurations.
This was frustrating because certain problems occurred while I was making adjustments to scenes that were already close to completion.
I also accessed Maya through Omnissa Horizon for parts of the production, meaning that my workflow depended on both the software and the stability of the remote connection.
This sometimes made it more difficult to work efficiently, particularly when switching between scene adjustments, animation and rendering tests.
As a result, I became more cautious about making major changes to complex scenes and placed greater importance on maintaining recoverable versions.
Managing Late-Stage Changes
Another challenge was deciding how to approach adjustments after sequences had already been rendered.
During editing, I continued identifying opportunities to improve shot timing, camera movement and visual consistency.
However, even relatively small changes could require additional rendering.
This meant considering whether a modification would produce a noticeable improvement or introduce unnecessary risks and production time.
I began prioritising adjustments that directly affected the readability and pacing of the film.
For example, extending a shot could be more important than refining a minor background detail if it allowed the audience to understand the narrative more clearly.
This required a different approach from the earlier stages of production, where I had more flexibility to experiment with individual elements.
File Management and Recovery
Managing project files also became increasingly important as production progressed.
Working with multiple Maya scenes, textures, imported assets and rendered image sequences resulted in a large number of files and different versions.
At one point, I encountered a situation where a Maya scene appeared to be missing.
Fortunately, I was able to recover the file, but the experience made me more aware of the importance of maintaining backups.
Working across different computers and using cloud synchronisation also required additional attention to file locations and accessibility.
I needed to ensure that the correct versions of scenes and supporting assets were available when continuing work.
This experience demonstrated how file management could directly influence production efficiency, particularly when dealing with large and technically demanding projects.
Learning to Prioritise
As the film approached completion, I became increasingly aware that not every technical or visual issue could receive the same amount of attention.
Throughout production, I had sometimes spent considerable time attempting to refine relatively minor details.
Although this helped me develop my technical skills, I also began recognising that some adjustments had a greater impact on the final film than others.
I therefore started evaluating changes according to their contribution to the narrative and overall visual quality.
This was especially important when deciding whether to repeat a render, extend an animation or continue experimenting with a visual effect.
Rather than attempting to perfect every individual element, I began focusing more on the consistency and effectiveness of the film as a whole.
Reflection
The final production stage required me to balance technical refinement with the practical limitations of completing an independent animated film.
One of the main lessons was learning to recognise when additional experimentation was useful and when it risked delaying more important work.
The combination of software instability, rendering requirements, file management and late-stage adjustments encouraged me to develop a more organised and selective approach to production.
Although not every visual or technical element reached the level of refinement I initially imagined, these experiences helped me understand the importance of making decisions within the context of the complete project.
The most valuable outcome was developing a more realistic understanding of how creative ambition, technical limitations and production time influence one another when bringing an animated film towards completion.