1. Project Overview

Project Title: The Thread of Fate

1 min 55 secs surreal VFX short film

A symbolic exploration of probability, destiny, social hierarchy, identity, and philosophical individuality, expressed through stylised characters, atmospheric environments, and a psychological narrative structure.

Conceptual Foundation

The core idea is rooted in a reflection inspired by the Myth of Sisyphus and existential philosophy.

“Regardless of the highest highs or lowest lows, the more absurd the world becomes, the more one can – through moments of defiance – shape oneself into who one longs to be.”

Two symbolic archetypes emerge:

• Mud — rough, imperfect, rooted in hardship, but resilient
• Porcelain — flawless, elevated, refined, but fragile

These are not merely materials but psychological archetypes of human experience and social identity.

Final Video:

Story Premise

A first-person POV is pulled into a labyrinth woven with symbols of fate. Along a glowing red thread, two contrasting worlds unfold:

• Porcelain collapses under the weight of perfection
• Mud strengthens through persistence

At the climax, the POV is presented with a choice: save one, only one.

Instinctively, many would save Mud – the one who appears stronger, more determined, more “deserving.”

Yet within the story, Mud chooses differently:

Mud reaches out to save Porcelain.

Not out of obligation, superiority, sympathy, or societal hierarchy, but from an inner conviction that origin does not define worth.

This gesture becomes:

• a rejection of deterministic fate (breaking free from the red thread),
• an affirmation of agency amid absurdity,
• and a subtle critique of secular value systems that decide who is worth saving.

Philosophical Lens

This project reflects on how identity is formed and perceived:

The self you see is not your true self. The self others see is not your true self. The way you see others may reveal who you truly are.

Through collapse, convergence and transformation, the film asks how individuals define themselves when confronted with randomness, societal hierarchy, and existential choice.

Making of Video:

2. Pre-Production: Concept and Story development

2.1 Conceptual Research

Absurdism & Character Archetypes

The first concept derived from The Myth of Sisyphus—a figure pushing uphill—naturally led to a mud-covered lower-class character.

To contrast this, I explored metaphors for privilege: porcelain, glass, polished surfaces, which visually convey refinement and pressure.

To avoid tying the characters to specific gender, race, or age, I initially chose two hands as symbolic representatives. This ensured the focus stayed on choice, fate, and inner spirit, rather than appearance.

However, during storyboarding, I realised:

• hand gestures lack emotional range
• full-body movement (especially for dance and collapse) was necessary

Thus I designed sculpture-like bodies—non-gendered, minimalist, and expressive.

Identity, Class Distinction, and Environment

Since the narrative is tied to a modern city:

• the lower-class world became a claustrophobic underground base, symbolising social pressure from above
• the upper-class world became an ornate high-rise interior inspired by European palaces

This grounded the hierarchy in a recognisable contemporary context.

Fate vs Choice & Symbolic Movement

To express emotional and thematic contrast:

• Porcelain destroys objects – symbolising cynicism, pressure, loss of control, and internal fragility
• Mud performs ballet, inspired by Degas’ dancers – representing discipline, aspiration, and resilience emerging from an unlikely place

This decision aligns both visually and conceptually with the project’s core themes: upward striving, contrast in class expectations, and the confrontation of fate.

2.2 Concept Development

Early brainstorming keywords shaped the symbolic vocabulary:

• Probability: poker, chess, pool balls, dice, roulette
• Destiny: red thread, clue wall, clock, photographs, pathways
• Upper class: books, piano, wine, polished surfaces, chandeliers
• Lower class: stone, mud, rubbish, ruined city
• Character materials: glossy porcelain, cracked statue, rusty metal, wet soil, hill/underground imagery

To visualise the evolving environment and emotional tone, I used AI tools (ChatGPT image generation) to test compositions. This revealed a practical issue: hand-only characters lacked expressive range, especially in scenes requiring gestures, collapse, or dance. The visual tests confirmed the need for full-body silhouettes – simple but expressive.

These early sketches, AI tests, and compositional notes guided my refinement of the scenes and pacing.

2.3 Storyline

The storyline went through three major versions:

• Version 1: literal Sisyphus stone-pushing
• Version 2: mud vs porcelain hands
• Version 3 (final): full-body porcelain and ballet-mud characters

The narrative evolution involved:

• Gathering references (paintings, cinematography, cityscapes)
• Visualising transitions between worlds
• Testing symbolic continuity (dice → maze → collapse)
• Discussing with Manos and Emily, who advised making emotional cues more readable and reducing unnecessary abstraction
• Rebalancing both character arcs to ensure they contrast yet parallel each other

Below is the final storyline that emerges from these refinements.

2.4 Previs, Animatics, and Storyboard

Because the narrative included five distinct environments and non-linear transitions, a 2D storyboard alone was insufficient, particularly for camera motion, floating sequences, and world collapse.

Therefore, I created:

This helped refine timing, continuity, and symbolism before production.

3. Pre-Production: Audio-Visual Research

3.1 Visual Research & References

Scene Reference

Reference for S1 – Opening (Clue Wall)

I referenced suspense-driven visual language from contemporary mystery dramas (because popular and successful TV dramas over the last 10 years have most frequently featured suspense topics). The goal was to create an immediate sense of intrigue while introducing the two characters through a symbolic clue wall.

Reference for S2 – Transition Space (Dreamlike Maze Entrance)

This scene draws on dreamlike compositions to heighten the feeling of uncertainty and randomness – establishing the thematic link to probability and destiny.

Reference for S3 – Upper-Class Room

I also studied interiors from:

• Madrid Palace
• The Louvre’s historic rooms
• The Kew Palace at Kew Garden in London

Common features: leisure rooms, musical instruments, oil paintings, polished floors, and chandeliers.

Elements such as chess sets, poker cards, chandeliers, and ornate décor help visualise a luxurious yet pressured high-class environment.

S4 – Lower-Class Underground Scene

For the lower-class environment, I researched:

• Urban underground spaces in China, Japan, and London

Common features: compressed tunnels, pedestrian footbridges, cluttered alleys, rusty ground textures.

This led to a dark, crowded, cluttered world, surrounded by towering city structures, with scattered newspapers and broken bottles. The Ballet Dancers reference appears on a worn newspaper, symbolising inspiration emerging from debris.

The focus was on a darker, crowded, and worn environment inspired by urban underground spaces. Grass and small plants were included deliberately to symbolise growth and resilience emerging from harsh conditions.

S5 – Climax (Cracks, Chaos, Clocks)

References included surreal compositions featuring suspended objects and strong upward lighting. Clock designs and floating character poses helped construct the mid-air convergence and temporal symbolism.

S6 – Closing (Return to Clue Wall)

Uses the same references as S1, with modifications to the characters’ photographs. No additional source reference was required, as the scene echoes the visual structure of the opening.

Scene Structure (Final Output) & Symbolism

S1 – Clue Wall (Opening): Suspenseful tone, introducing both characters.

S2 – Maze Entrance: A surreal transition, representing birth randomness and destiny.

S3 – Upper-Class Room: Luxurious, symbolic of privilege; includes paintings and polished objects.

S4 – Lower- Class Environment: Dark, compressed environment symbolising perseverance.

S5 – Convergence & Collapse: Chaos, floating debris, clocks symbolising asynchronous worlds, timelessness.

Specifically,

The S5 climax begins with cracks tearing through both worlds, releasing debris, light, and chaos. As gravity collapses, the two characters rise into mid-air, finally converging. This moment reveals the deeper connection between the POV and both characters – three timelines crossing in a single surreal space.

The three clocks in this scene symbolise the temporal logic of the story:

1. Two clocks at the front → represent the separate worlds and personal timelines of the two characters. Each exists in its own world and rhythm.
2. The third clock in the background → remains completely still at the start of the meeting. Its frozen state represents a neutral, surreal time zone – a “time gate” where the two worlds can overlap. Time does not flow here unless a significant choice is made.

When the POV makes the choice, and the lower-class character reaches for the upper-class character, the meanings of all three clocks shift:

• The front clocks stop, symbolising the dissolution of fate in both individual worlds.
• The back clock begins to move for the first time, marking the moment the two characters finally transcend their predetermined paths.

This temporal shift completes the act of salvation:

a moment where time breaks, reconnects, and begins again, not according to fate, but according to choice.

S6 – Final Clue Wall Update: similar to S1 but with altered photos.)

S6 returns to a setting visually similar to the opening clue wall, creating a circular narrative structure. However, several details reveal the transformation that has occurred:

• The photographs have changed
• A sheet of cracked glass now lies in front of the clue board, symbolising the aftermath of both worlds colliding. The fracture echoes the shattering of fate and perception witnessed in the climax.
• Directional lighting and strobing highlights simulate the feeling of an investigator’s flashlight sweeping across the surface, as if the POV has returned to examine the final state of the two lives after the surreal encounter.

Together, these elements signal that although the world has returned to its initial perspective, the understanding of both characters – and of fate itself – has fundamentally shifted.

Symbolism from Paintings

As I was doing the Gaussian Splatting topic for my final thesis, I studied paintings and found several works relating to hierarchy, vanity, time, and existential reflection:

• The Ambassadors – depth of perception, mortality, distorted truth
• An Allegory of Prudence – authority and temporal judgement
• Vanitas Still Life – impermanence and moral critique
• Degas’ Ballet Dancers – discipline, aspiration amid constraint

I integrated these symbolically:

• Allegory of Prudence appears in the high-class room (S3), reinforcing the inherited hierarchy.
• Ballet Dancers appear in the underground scene (S4), representing aspiration beneath oppression.

Red Thread Symbolism

The red thread is the structural spine of the film. It symbolises:

• Fate
• Connection
• Control (POV’s agency)
• The path through randomness

Its eventual breaking reflects the characters’ rejection of deterministic destiny.

3.2 Cinematic, VFX, and Audio Research

To refine atmosphere and emotional tone:

• I researched cinematic lighting using volumetric fog and directional spotlights to highlight symbolic objects.

Gaussian Splatting Research

Alongside my final thesis, I explored whether Gaussian Splatting (GS) could be integrated into this project’s visual pipeline.

During experimentation, I tested:

• 3D GS scans of paintings from the National Gallery
• City-scene captures around London
• A full GS reinterpretation of The Ambassadors for my thesis work

Through these tests, several limitations became clear:

• Lighting inconsistency produced unstable or noisy reconstructions.
• Stylised scenes did not translate well through GS due to its dependence on physically accurate capture conditions.
• For independent creators, GS remains technically and computationally demanding. Without engineer-supported machine learning models, the current open-source GS tools cannot deliver predictable, production-ready results.
• Current industry adoption of GS is still mainly limited to previsualisation, lighting studies, and exploratory workflows.

These findings are supported by research and industry commentary:

Hodge, S. (2025). NeRFs, Gaussian Splatting and the Future of VFX. Available at: https://www.fxguide.com/fxpodcasts/nerfs-gaussian-splatting-and-the-future-of-vfx/ (Accessed: 12 November 2025).

Silva Jasaui, A., Ferreira do Amaral, J. and Mendes, D. (2024) ‘Virtual Production: Real-Time Rendering Pipelines for Indie Studios and the Potential in Different Scenarios’, Applied Sciences, 14(6), pp. 2530: 1-21. Available at: https://www.mdpi.com/2076-3417/14/6/2530 (Accessed: 12 November 2025).

Garage Farm (2024) Volume VFX: Innovations in Virtual Production Technology. Available at: https://garagefarm.net/blog/volume-vfx-innovations-in-virtual-production-technology (Accessed: 12 November 2025).

Given that my project had already progressed into previs and animatic development, and I achieved more reliable and aesthetically controlled results through manual modelling, texturing, and lighting, I decided not to integrate GS into the final film.

Thus, Gaussian Splatting remained an important part of my research process, but not the final production pipeline.

Audio Research

Due to copyright constraints:

• I attempted to generate a playable version from sheet music
• But the computer-rendered performance lacked the emotional nuance needed

Eventually, I generated an original BGM using Suno AI, iterating more than twenty times and using descriptors inspired by the above two tracks. Although not perfect, it harmonises with the film’s emotional arc and pacing.

3.3 Asset Planning

I created an asset list specifying:

• Models required per scene
• Animation requirements
• VFX needs (wine shatter, thread behaviour, volumetrics, clock rotation, debris simulation)
• Rough frame counts for each shot

This served as the foundation for my production schedule and rendering plan.

4. Production process

4.1 Modelling

Character Modelling

Base meshes for both charaters were generated using Meshcapade Me https://me.meshcapade.com

The raw meshes’ proportions were edited on the website: the high-class character was made slightly taller and more physically defined, while the lower-class character remained slimmer and more grounded.

I sculpted additional anatomical details and added subtle facial and body refinements.

Hair was added only to the high-class character to emphasise material privilege. In contrast, the lower-class character remains without hair, symbolising minimalism and deprivation.

Texture Development

Both characters required extensive shader experimentation in Blender:

• Mud Character A black, rusty procedural texture was combined with subtle emission to create a “glow beneath the dirt,” referencing resilience and inner vitality. Achieving this balance took several trials to avoid over-brightening the dark environment.

• Porcelain Character I developed two texture forms:
  1. Pristine glossy porcelain (for the beginning)
  2. Cracked fragile porcelain (after the collapse)
  Noise and Voronoi patterns were layered to simulate ceramic imperfections. Gold flecks and thin metallic lines were introduced to convey nobility and the illusion of refinement.

To create the fractured version, I used Cell Fracture on a duplicated mesh. This required significant mesh cleanup- removing irregular points, reducing poly count, and fixing broken normals – before the model could successfully undergo rigging and deformation.

Rigging & Retargeting

Because the Meshcapade-generated bodies had no usable rig, I had to:

• Manually rig the base meshes
• Retarget animations for each scene
• Re-rig fractured versions separately

This took considerably longer than expected, as each variation required adjustment to maintain clean deformations during choreography and mid-air motions.

Scene Modelling

S1 – Opening (Clue Wall)

Everything was modelled from simple planes in Blender: photos, strings, notes, pins, layered paper textures, and the cracked glass for the final scene.

S2 – Maze Transition

The visual design draws heavily from Brutalist architecture – raw concrete walls, minimalistic doors, sharp edges – representing the ambiguity of birth and probability.

Additional symbolic elements include:

• Poker patterns (A – 10 were assets from www.freepik.com, K/Q/J/Joker custom generated via prompts with ChatGPT; textures edited in Blender)
• Human evolution silhouettes (also generated via ChatGPT and then remade as Blender meshes)

• A spiralling path created by animating these silhouettes as a moving ribbon inside the tunnel

This scene was intentionally simple in geometry but rich in conceptual symbolism.

S3 – Upper-Class Room

Only the piano and chandelier were free online assets to speed up production and enhance the room’s visual richness.

Most assets – doors, walls, columns, window frames, furniture, pokers, wine, glass, chess, books, table, curtains (cloth simulation) and painting frames – were modelled from scratch.

S4 – Lower-Class Underground City

High-rise buildings assets are from free Kitbash 3D and Max Hay‘s buildings assets.

To emphasise the hierarchy gap, I created additionally:

• A pedestrian footbridge crossing the centre of the scene (symbolic separation)
• Layered lighting to keep the focus on the character
• Debris, litter, rust, and uneven roads

Plants like grass and small flowers were manually placed to represent growth emerging from hardship.

S5 – Falling / Collision Space

The mid-air surreal environment, including floating clocks, debris, fragments, and volumetric fog, was fully modelled and shaded in Blender.

Most assets were created from simple planes that were subsequently deformed, fractured, or instanced.

S6 – Final Clue Wall

Similar to S1, but with:

• Updated character photos
• A fractured glass panel (created using Cell Fracture)
• New lighting and dust volumes to signal aftermath

4.2 Animation

Character Animation

Both characters required extensive rigging work in Blender, especially the upper-class character with its two distinct forms (glossy intact and fractured).

Because the base meshes were generated from online tools and later reshaped during texturing, each rigging pass involved:

• rebuilding skeletons
• remapping meshes
• correcting distortions
• repeatedly adjusting weights to match the stylised textures

The fractured version of the upper-class character was particularly time-consuming. After applying Cell Fracture, a large amount of manual cleanup was needed – removing irregular geometry and stabilising mesh topology – to ensure the model could deform properly in animation.

Most humanoid movements were sourced from Mixamo, then refined in Blender.

For the Mud character’s ballet performance, I used motion-capture data exported from Meshcapade Me, allowing a more expressive and grounded animation.

However, the floating, falling, and mid-air collision shots had no suitable, ready-made assets, so:

• I adapted existing Mixamo clips
• then manually edited them frame-by-frame to achieve believable motion in zero-gravity and chaotic environments

Object Animation

Red Thread / Rope

Created using curve paths and animated via path position.

A custom rope-like material was built to maintain the physical feel of a woven thread.

Wind, particles, and atmospheric motion

Subtle turbulence and particle systems were added to the S1 clue wall scene to enhance mystery and depth.

Dice

Dice modelled and keyframed manually

S2 Spiral Tunnel

The ascending spiral inside the maze was built with Geometry Nodes, adapting the logic of staircase construction and replacing steps with:

• human evolution silhouettes
• poker cards

These elements swirl around the red thread to symbolise probability and destiny.

Poker Cards

Pokers were modelled from scratch.

Face designs (K, Q, J, Joker) were regenerated using ChatGPT for a sarcastic, statue-like aesthetic.

Their scattering behaviour in S2 was driven by a particle system, animated in three separate phases.

Clocks (S5 & S6)

Clock hands and rotations were animated manually frame-by-frame for full control over the symbolic timing across all three worlds.

Camera Animation

Camera motion begins in first-person POV, guiding the audience through the characters’ stories.

Throughout the film, cameras were choreographed to emphasise emotional contrast through:

• slow close-ups (internal conflict)
• wide shots (environmental hierarchy)
• drifting transitions (dreamlike shifts between scenes)

During the maze transitions, dice-rolling moments, and the falling sequence, I applied deliberate focus shifts and soft arc movements (following Manos’ advice). This helped create a smoother, more surreal flow while reducing the appearance of uneven animation.

Because of rendering constraints – and a laptop crash mid-production – the final camera timings were tightened. I prioritised:

• efficient camera paths
• reduced redundant movement
• fewer but more meaningful keyframes

This allowed me to complete all renders on time while maintaining:

• reduced render load
• remain narrative clarity
• the full abstract storyline within the deadline

However, this optimisation also resulted in a few unnatural or abrupt transitions that could only be partially corrected during post-production. Despite these compromises, the overall visual flow remains coherent and consistent with the film’s surreal tone.

4.3 VFX and Simulation

Paper Burn Effect – S1 Opening

The “burning photo” transition in Scene 1—where the framed picture burns away to reveal the entrance to the maze—was created entirely using Blender’s Shader Editor.

I built the effect using:

• Noise textures to drive the irregular burn pattern
• Emission shaders to simulate glowing, burning edges
• Alpha mapping to gradually dissolve the paper into ash

This allowed the photo to feel as though it is naturally burning away from the inside, revealing the surreal world behind it without relying on external simulations.

Water Reflection – S2 Maze Scene

The water surface in S2 was created entirely using procedural shaders in Blender.

Animated noise textures controlled subtle distortion, producing a realistic ripple across the pool.

In the reflection, I placed a silhouette of a boy standing at the water’s edge, symbolising the POV observing his own fate from outside the maze.

To enhance realism, I carefully matched the coloured window reflections on the water surface—taking inspiration from the lighting inside the Alhambra Generalife. This required precise alignment between window glass shaders and the water shader’s reflectivity to achieve the correct hue and light behaviour.

Liquid Simulation – S3 Breaking Wine Glass & Wine Splash

The wine simulation consisted of two separate fluid simulations:

1. Houdini Simulation
   • Used for the mid-air splash created when the glass is thrown.
   • Houdini was chosen because its FLIP solver handles free-form fluid motion with higher stability.
   • The simulation was exported as an USD file and imported into Blender.

2. Blender Simulation
   • Used for the secondary splash when the wine hits the chessboard.
   • Simulated directly in Blender to avoid heavy cache sizes and RAM overload.

This separation was necessary because:

• Houdini simulations required large cache files that exceeded my laptop’s available storage and RAM.
• Blender’s scene already contained multiple force fields, and running two full fluid passes inside Blender would have destabilised the setup.

However, the workflow was further complicated when my laptop crashed mid-production. Switching to different university machines required re-caching, and Blender’s fluid caching inconsistently altered the simulation origin. This forced an additional re-simulation phase, resulting in a good but not the original ideal version. Considering time and hardware limitations, I had to use this most viable outcome.

Plant (Sprout) Growth Simulation – S4 Lower-Class Scene

The flowers were modelled manually, then combined with grass assets to form the base of a growth simulation.

Using Blender’s Particle System with hair dynamics, I animated the flowers to emerge gradually from the ground, symbolising resilience and hope.

The terrain (the ground under the feet) used layered noise textures for a rusty, moist, underground surface, enhancing the contrast between decay and life.

Cracks, Fracture Simulation & Light Wave — S4 & S5

Cracks and fractured debris were created using the RBDLab plugin in Blender:

• Crack paths were first drawn directly in the viewport using annotations
• Used RBDLab to generate fracture patterns along these paths
• Parameters such as shard size, debris amount, and dust particles were adjusted for dramatic effect

Additional smoke, dust particles, and floating debris were simulated to enrich the collapse sequence.

Light Wave Effect

The light-wave effect was built in Blender as well, using an asset I originally developed during my exploratory practice (“Disposable Camera”). It relies on a combination of geometry nodes and shader-driven emission to shape the wave, control its edge distortion, and animate intensity falloff. For this project, I adapted the effect to a larger scale and faster dispersion to convey the wind pressure and ground shockwaves as the city collapses.

Due to my laptop crashing before final lighting and camera refinements, the final render did not fully match the higher-quality previews I achieved earlier. Switching between university machines and dealing with unexpected remote-computer disconnections made it difficult to maintain consistent camera angles and lighting setups. As a result, the crack and light-wave scenes look slightly less polished than the original simulations, though the core visual intention remains clear.

4.4 Lighting

Each scene required a different lighting strategy, but the overall goal remained consistent:

a cinematic opening, a dreamlike transition, clear character focus, and an intensified mood in the S5 collapse scene.

S1 & S6 – Clue Wall / Ending Wall

Lighting these scenes was relatively straightforward:

• Standard three-point lighting to reveal the clue wall.
• Added volumetric haze to introduce atmosphere and a soft cinematic mystery.
• S6 follows the same structure but with adjusted intensity to highlight the changed photographs and cracked glass.

S2 – Maze Entrance / Reflective Pool

This was one of the most challenging lighting setups because the light needed to:

• Correctly color the stained-glass windows,
• And accurately reflect those colors onto the water surface.

To achieve this, the following three points have to be in the same scene:

• A Blender sky texture for base ambient light.
• A back spotlight behind the window and the boy (POV reflection).
• A second directional spotlight to illuminate the tunnel door in the mid-background.

Balancing the three allowed the water surface to capture realistic reflections while keeping the path readable.

Inside the tunnel itself, the lighting was built from two contrasting spotlights—orange and purple-blue—positioned opposite each other. This created a dreamlike gradient and supported the transition into the surreal maze environment.

S3 & S4 – Upper-Class / Lower-Class

Initially, I lit the whole environment evenly to assist modelling, but this resulted in:

• No focus hierarchy
• Overexposed objects
• Characters blending into the environment

After receiving feedback from Emily, I reapproached the lighting completely:

1. Converted all geometry to white meshes and rendered the scenes as lighting-only drafts.
   • This removed texture distractions and exposed lighting errors clearly.

2. From the white-mesh passes, I identified:
   • Over-lit background areas
   • Missing rim light on characters
   • Shadows that flattened the scene
3. Re-lit both scenes with strict priorities:
   • Key light on the character
   • Secondary highlights on symbolic objects
   • Controlled darkness to preserve hierarchy and guide camera focus

This iterative process significantly improved clarity and mood.

S5 – Collapse / Mid-Air Convergence

Lighting in S5 was designed to create a surreal, collapsing atmosphere while ensuring the two characters remain the emotional focal point and the background clocks stay clearly readable. Key techniques:

• Soft rim lights on both characters to give them a sanctified, almost transcendent presence as their worlds collide.
• Volumetric fog and drifting particles to mimic falling dust, deepen the sense of collapse, and add a cinematic atmospheric layer.
• Directional glancing highlights on the floating clocks to maintain spatial clarity and guide the viewer’s eye through the chaos.

4.5 Rendering

Before the laptop crashed, I had successfully optimised my render time from 1 minute per frame down to around 24 seconds avg in Blender, without sacrificing visual quality. This improvement came from carefully adjusting:

• indirect light bounces
• noise threshold
• camera culling margin
• resolution and sampling
• persistent data
• various other Cycles optimisation settings

Ironically, this optimisation helped reveal the hardware problem: when re-rendering the same frames, the time suddenly increased from seconds to over 50 minutes, and the CPU repeatedly shut down due to overheating.

Due to these hardware issues and the instability of both my laptop and the university’s remote computers, I had to reconfigure my render settings again. This meant compromising:

• lowering certain visual details
• reducing resolution in selected scenes
• simplifying lighting passes

The priority became ensuring all frames could be rendered on time for compositing before the deadline.

The later rendering stage was stable, but this compromise introduced noise, colour shifts, and lighting inconsistencies, which I needed to correct during post-production.

5. Post-Production, Compositing and Video Editing

I used green screen footage to capture my real hand for pointing gestures and a separate falling action.

Only the hand footage was used in the final film (Scene 5).

The falling shot was originally intended for the ending, but due to the laptop failure and the need to work via remote machines at Uni, I decided to replace it with cracked-glass visuals and flash lighting, which communicated the idea more effectively.

Green Screen and Compositing were done in Nuke, after which I exported everything in Rec.709 and moved to DaVinci Resolve for motion blur, final colour grading and editing.

I chose DaVinci Resolve because:

• It runs more reliably on unstable hardware compared to Nuke
• Its colour management is more intuitive for matching shots
• It includes the Blender Cinematic LUT, which aligns perfectly with my Blender rendering pipeline

Colour grading was essential because each of the six scenes was rendered with slightly different tones.

Using DaVinci allowed me to unify exposure, contrast, and colour temperature across the whole film.

The text animation “Save one, Only one” was also created in DaVinci, using custom animation curves so the characters pass through the text before it pops forward into focus.

After final grading, all shots were exported from DaVinci and brought into CapCut for:

• Sound design
• Background music
• Fine-tuning animation curves

This final pass ensured that visual rhythm and audio timing matched the emotional pacing of the film.

6. Reflection

What Worked

The two distinguished character designs – the lower figure’s glowing resilience and the upper figure’s fragile porcelain elegance – successfully embodied the thematic contrast between class, identity, and fate. Their visual languages remained clear throughout the film and communicated the emotional and philosophical tension I intended.

The lighting revisions, especially the white-mesh relighting process, significantly improved atmosphere, hierarchy, and character focus across all scenes. Before my laptop failed, I also managed to optimise render times from one minute to about 24 seconds per frame, which allowed fast iteration and stronger visual development during early production.

The climax sequence, with collapsing worlds and the interplay of three clocks, conveyed the symbolic logic of fate, time, and choice as planned. This was one of the scenes where the intended meaning translated directly into the final visuals.

What Didn’t Work

The most significant setback was my laptop’s sudden failure in the middle of the whole production. This directly affected shot refinement, animation polishing, and rendering quality, forcing me to reduce settings and accept higher noise, colour inconsistencies, and lower-resolution outputs.

The hardware failure also meant I could not render AOV passes, something I regret, as it limited the accuracy and flexibility of colour grading and compositing.

Some camera transitions ended up more abrupt due to the need to shorten shots for render deadlines. The wine simulation had to be rebuilt on multiple machines, producing a less convincing result than the original. Altogether, these technical constraints reduced the level of visual refinement I had intended for the final cut.

How Research Shaped the Project

The research process directly informed every creative decision.

Camus and the Myth of Sisyphus shaped the existential tone and inspired the lower-class character’s quiet resilience. Allegorical paintings such as An Allegory of Prudence, Vanitas Still Life, and The Ambassadors influenced visual language around hierarchy, mortality, and perception. Degas’ ballet studies reframed perseverance as disciplined movement instead of physical labour. Concepts from probability and randomness guided the dice, maze, and spiral motifs.

Even though Gaussian Splatting was not included in the final production, studying its strengths and limitations helped shape my early visual exploration and spatial thinking.

I also experimented with stylised filters – painting-like and manga-inspired looks generated through Blender’s compositor. In the end, I chose not to apply them broadly, as they masked the detailed textures and material qualities I had carefully developed for the characters and environments. However, I retained a subtle painterly effect in the S2 entrance sequence to enhance its dreamlike transition.

What I Learned

This project strengthened my understanding of shader creation, fracture simulation, geometry nodes, cinematic lighting, and general pipeline management. I learned how to balance abstraction with emotional clarity, and how to maintain creative direction even when hardware failure disrupted the pipeline. It also reinforced my ability to merge technical VFX processes with philosophical storytelling.

Future Improvements & Potential

With stable hardware, I would re-render each scene with full AOV passes, refine lighting consistency, rebuild the wine simulation, and polish camera transitions.

The project itself holds future potential—either as an extended surreal short or as a multi-screen installation exploring fate, hierarchy, and perception through spatial storytelling.

Despite the challenges, the film still conveys how personal choice can reshape destiny and how inner spirit ultimately defines one’s value.