High-End Post-Processing Architecture for Real-Time Abstract Graphics

Full report

On this page

Introduction to Premium Abstract Post-Processing

The creation of dreamlike, mesmerizing real-time abstract graphics requires an architecture where post-processing is elevated from a secondary corrective measure to a primary artistic medium. In procedural and abstract rendering, the underlying geometry frequently serves merely as a scaffold. The true perceptual richness of the image is defined by the complex interplay of lighting, shading, volumetric phenomena, and the simulated physics of the virtual camera lens. However, the most significant risk in high-end post-processing is the destruction of geometric legibility. Excessive bloom, poorly calibrated chromatic aberration, or overly aggressive tone mapping can rapidly devolve a highly structured mathematical form into a chaotic, indistinct gradient that strains the viewer's eyes and destroys the composition.

To prevent post-processing from overwhelming the underlying geometry, a modern pipeline must operate on strict physical principles while allowing for non-physical, artistic deviations in a highly controlled manner. This necessitates an architecture that operates exclusively in linear light for all optical simulations, carefully manages the dynamic range of saturated procedural colors, and utilizes edge-aware or depth-aware masking to preserve high-frequency structural details. Advanced post-processing pipelines for premium real-time applications must seamlessly integrate multi-scale bloom, precise optical lens distortions, physically derived depth of field, and robust temporal accumulation systems. The synthesis of these elements, governed by advanced color management and sophisticated tone mapping algorithms, dictates the final atmospheric depth and continuity of the rendered scene.

Color Management, Displays, and the Browser Ecosystem

The foundational layer of any high-end graphics pipeline is its color management architecture. Modern real-time platforms, particularly those operating within web environments via WebGPU or native rendering APIs, face a highly fragmented ecosystem of display capabilities, transfer functions, and gamut representations. Operating in a linear color space is a non-negotiable prerequisite; light accumulation, convolution blurring, and physical lens simulations are mathematically invalid in gamma-encoded spaces.

Linear Light, sRGB, and Gamma Correction

In a physically based rendering environment, the pipeline accumulates light in linear space, meaning a pixel receiving twice as much light possesses exactly twice the numerical energy value. However, physical display hardware expects a non-linear signal, typically governed by the sRGB transfer function or the Electro-Optical Transfer Function (EOTF) of a specific standard. The Opto-Electronic Transfer Function (OETF) applies an approximate [Figure omitted from source export] curve to the linear data to compress it for standard displays1.

When procedural imagery is generated with extreme saturation and brightness—a common occurrence in abstract art—evaluating these values in linear space ensures that energy is rigorously conserved when light bounces or scatters. If a post-processing effect, such as bloom or depth of field, is applied after the gamma correction step, the brightest areas will not scatter correct amounts of light, resulting in muddy, unnatural gradients that destroy the illusion of glowing geometry. The entire post-processing chain must execute before the final OETF is applied.

Wide-Gamut, HDR, and WebGPU Integration

Modern hardware frequently supports wide-gamut color spaces, such as Display P3, and High Dynamic Range (HDR) outputs capable of pushing thousands of nits. In WebGPU, the configuration of the canvas context dictates how the browser compositor handles the swapchain textures2. Requesting the display-p3 color space allows the application to output colors that fall outside the traditional BT.709 sRGB triangle, enabling much deeper, more saturated procedural reds and greens that are critical for mesmerizing abstract aesthetics2.

Furthermore, HDR display routing in WebGPU involves the toneMapping property. Setting the mode to extended allows the canvas to bypass the standard standard dynamic range (SDR) clamping, transmitting values brighter than reference white to the display3. When rendering premium abstract graphics, this extended range is crucial for true volumetric glow. However, it introduces significant complexities regarding hardware capability mapping. Depending on the underlying graphics API (Vulkan, Metal, or DX12), the surface capabilities expose varying extended-range formats, such as ExtendedSrgbLinear or ExtendedDisplayP34. If the display lacks native HDR capabilities, the browser or operating system will apply a fallback tone mapping curve that the developer cannot control, often leading to unpredictable crushing of highlight detail4. Therefore, the rendering pipeline must query device capabilities and gracefully fall back to an internal SDR tone mapping operator if the hardware cannot natively support the extended dynamic range.

Mobile OLED Behaviors and Black Smearing

Mobile platforms present unique optical and hardware challenges, particularly regarding OLED displays. OLED technology allows for theoretically infinite contrast ratios because individual pixels can be turned off entirely to achieve perfect black. However, this introduces a hardware-level artifact known as "black smearing." Black smearing occurs because OLED pixels require significantly more time to transition from an entirely powered-off state to an active, illuminated state than they do transitioning between two differing active states.

In dark, abstract scenes featuring floating glowing geometry against a black void, camera or object motion will cause the geometry to leave prominent purple or black trailing artifacts. To mitigate this at the post-processing level, the pipeline must artificially elevate the black floor by adding a minuscule baseline value (e.g., [Figure omitted from source export] in linear space) or applying a subtle layer of film grain. This ensures the OLED pixels remain active and voltage-primed, circumventing the hardware-level smearing while maintaining a perceptually pure black background to the human eye.

Exposure Control, Tone Mapping, and Color Grading

In highly saturated procedural imagery, tone mapping is the most critical stage for defining the final aesthetic. Because procedural shaders often output colors far exceeding the normalized [Figure omitted from source export] range, these values must be mapped back to the display's capabilities without destroying hue integrity or contrast.

Exposure Control via Exposure Values (EV)

Exposure control must mathematically precede all tone mapping operations. Operating akin to a physical camera, exposure is applied as a scalar multiplier to the linear scene data. Utilizing Exposure Value (EV) units allows for intuitive, logarithmically scaled brightness adjustments6. In EV systems, an increase of [Figure omitted from source export] EV corresponds to a doubling of the accumulated luminance, mirroring the perceptual response of the human eye6. The conversion between a user-defined EV and an internal linear multiplier is formulated as [Figure omitted from source export]6. This allows the post-processing pipeline to gracefully handle scenes ranging from dark, moody voids to overwhelmingly bright procedural suns without altering the underlying shader logic.

Soft-Knee Thresholding and Gamut Mapping

Before the primary tone mapping curve is evaluated, a soft-knee compression algorithm can be utilized to gracefully transition high-dynamic-range values into the tonemapper's optimal working range8. A soft knee defines a threshold below which colors are completely untouched. Above this threshold, a quadratic or exponential curve smoothly rolls off the brightness, preventing hard clipping of out-of-gamut colors8. This is critical when preserving the luminance ratios of saturated colors that approach the boundaries of the display's gamut11.

Operator Comparisons: Reinhard, Uncharted, ACES, and AgX

Different tone mapping operators handle the roll-off (the shoulder), the darks (the toe), and the preservation of chrominance in vastly different ways, fundamentally altering the artistic outcome.

The original Reinhard operator, defined mathematically as [Figure omitted from source export], compresses the entire luminance range uniformly. While it effectively prevents clipping, it severely diminishes midtone contrast, causing abstract graphics to look flat, gray, and desaturated14. A luminance-based Reinhard variant, which calculates the luminance of the pixel, applies the Reinhard curve, and then scales the original RGB channels by the resulting ratio, preserves color ratios better but still lacks the cinematic punch required for premium visuals16.

The Uncharted 2 operator, formulated by John Hable, relies on piecewise power curves to emulate the response of photographic film15. The mathematical foundation defines a curve with a linear section, a toe, and a shoulder, parameterized as [Figure omitted from source export]16. This operator allows technical artists to dictate exactly how shadows are lifted and how highlights roll off, maintaining contrast in the midtones while softly compressing overexposed glows17. It remains highly effective for abstract geometry where stark contrast between shadows and highlights is desired14.

The Academy Color Encoding System (ACES) provides an industry-standard look that yields high contrast and rich saturation. However, ACES suffers from notorious color-shifting issues when dealing with extreme, narrow-band procedural colors15. For instance, a pure, ultra-bright procedural blue will skew toward cyan and eventually shift to white under a standard ACES fit15. In abstract graphics, where a specific neon hue is aesthetically critical, this chromatic shift is often unacceptable.

AgX has emerged as a superior alternative for handling high-energy inputs gracefully1. AgX operates by applying a [Figure omitted from source export] inset matrix to pull colors toward the achromatic axis, followed by a [Figure omitted from source export] space encoding and a sigmoid function approximation1. After the curve is applied, an outset matrix restores the saturation. This prevents the primary color components from clipping at different rates, ensuring that a bright, saturated procedural red smoothly desaturates to pure white rather than awkwardly shifting to yellow or magenta15. Integrating AgX via GLSL or HLSL shaders allows for superior highlight handling in procedural environments where emissive values can exceed 100 nits1.

Alternatively, the Khronos PBR Neutral tone mapper was designed specifically to prevent any hue shifting, rolling off luminance while strictly preserving the input chromaticity20. This operator is invaluable for abstract scenes where absolute color fidelity and product-accurate hues are more important than achieving a dramatic, "cinematic" look22.

Color grading is then applied sequentially after tone mapping to impart stylistic shifts, utilizing 3D Look-Up Tables (LUTs) or American Society of Cinematographers Color Decision Lists (ASC CDL) to manipulate shadows, midtones, and highlights without breaking the dynamic range established by the tone mapper.

Bloom, Glow, and Convolution Architectures

Bloom is the cornerstone of the "glow" aesthetic in abstract graphics, simulating the scattering of light within a physical lens or the vitreous humor of the human eye. However, naive bloom implementations destroy geometric detail by washing the screen in low-frequency light. Premium pipelines rely on multi-scale, energy-conserving convolution architectures.

Separable Blur and Mip Chains

The traditional approach to generating bloom involves rendering the scene, thresholding the brightest pixels, and applying a Gaussian blur. Because a true 2D Gaussian blur of radius [Figure omitted from source export] requires [Figure omitted from source export] texture fetches, it is computationally prohibitive. It is therefore mathematically separated into two 1D passes—horizontal and vertical—reducing the complexity to [Figure omitted from source export]23. To achieve large, atmospheric glows, this separable blur is applied across a downsampled mipmap chain. However, even separable Gaussian blurs require significant texture fetching, which consumes immense memory bandwidth, particularly crippling on mobile devices or complex web scenes23.

Kawase Blur and Hardware Exploitation

The Kawase blur algorithm optimizes convolution by leveraging the GPU's hardware bilinear interpolation24. Rather than sampling every pixel in a large kernel, Kawase blur samples four corners at increasing distances from the current pixel across multiple passes, often executed at half or quarter resolution23. By incrementing the kernel offsets (e.g., 0, 1, 2, 2, 3\) across successive passes, the algorithm achieves a smooth, Gaussian-like distribution with a fraction of the computational overhead23. It effectively ping-pongs between two equally sized textures, accumulating a highly stable blur23.

Dual-Filter Blur Optimization

An evolution of the Kawase approach is the Dual-Filter blur, introduced by Marius Bjorge23. Unlike Kawase blur, the dual filter structurally integrates progressive downsampling and upsampling. It utilizes different spatial kernels for the downsample pass (sampling four covering pixels to smudge neighborhood information) and the upsample pass, effectively eliminating noise and temporal instability23. This technique yields a smooth, multi-scale bloom that perfectly captures wide, atmospheric dispersion while utilizing significantly less bandwidth than standard separable filters, making it the industry standard for high-end real-time bloom23.

Soft-Knee Bloom Thresholds and Energy Conservation

To prevent the entire image from becoming a hazy blur, a pre-filter pass applies a threshold to extract only the most luminous pixels before the downsampling chain begins. A hard threshold creates aggressive aliasing, temporal flickering, and blocky artifacts, so a soft-knee threshold is strictly required26. The knee math evaluates the maximum color component [Figure omitted from source export], and computes a contribution factor [Figure omitted from source export] where [Figure omitted from source export] is the threshold26. This bends the brightness curve downward smoothly, ensuring anti-aliased edges on emissive geometry remain intact before blurring26.

For a multi-scale bloom to remain physically plausible, the energy recombined during the upsampling phase must be normalized. If the accumulated bloom simply uses additive blending without balancing the weights of the mip levels, the center of the bloom will overexpose uncontrollably. A well-tuned multi-scale bloom uses fractional weights (e.g., [Figure omitted from source export]) combined with a slight dirt or smudge texture to simulate physical lens scattering and elevate the perceptual richness of the glow. Selective bloom, utilizing stencil masks or custom depth comparisons, further ensures that certain abstract geometries remain razor-sharp and unaffected while their immediate surroundings emit intense light.

Optical Aberrations and Lens Distortions

Simulating the imperfections of a physical camera lens breaks the mathematical sterility of real-time rendering, introducing organic depth and atmospheric continuity that grounds abstract geometry in reality.

Chromatic Aberration: The Accommodation-Vergence Conflict

Chromatic aberration (CA) occurs when a lens fails to focus all wavelengths of light to the same convergence point, causing color fringing along high-contrast edges. In post-processing, this is simulated by separating the red, green, and blue channels and scaling or offsetting them based on the distance from the screen center. While aesthetically popular, the indiscriminate application of lateral CA—shifting colors radially across the 2D image plane—causes severe eye strain and destroys line clarity27.

The human visual system relies heavily on sharp contrast edges for accommodation (the physical focusing of the eye's lens). When lines are artificially split into a rainbow gradient, the eye continuously attempts to bring the disparate wavelengths into focus, leading to visual fatigue and exacerbating the accommodation-vergence conflict, particularly in stereoscopic or highly immersive environments29.

However, chromatic aberration can drastically enhance depth perception when applied longitudinally, meaning the color separation is inextricably tied to the depth buffer and the circle of confusion29. Because the human eye naturally experiences chromatic aberration where blue light focuses at a different depth than red light, binding the CA radius exclusively to out-of-focus areas enhances the realism of the depth-of-field effect29. When objects fall out of focus, applying a slight blue-yellow or red-cyan fringe mimics natural dioptric power variations, enriching the atmosphere without straining the viewer's focus on the sharp, subject geometry33. Radial chromatic separation can still be employed, but it must be masked by the depth buffer to ensure that the focal plane remains perfectly sharp.

Barrel Distortion, Pincushion, and Vignetting

Barrel and pincushion distortions warp the UV coordinates of the image, curving straight lines outward or inward, respectively, mimicking the curvature of wide-angle or telephoto lenses. A subtle barrel distortion combined with radial blur implies a wide-angle lens, subtly wrapping the abstract geometry around the viewer and enhancing the sense of scale. To maintain geometric legibility and prevent nausea, this distortion must remain below a 2% displacement threshold.

Vignetting gently darkens the periphery of the image, drawing the viewer's eye toward the center while simultaneously hiding the stretching artifacts and texture clamping caused by barrel distortion at the screen edges. This effect is mathematically applied as a radial smoothstep function originating from the screen coordinates center, further mimicking the physical light falloff of a camera barrel.

Depth, Atmosphere, and Volumetrics

Abstract graphics demand a profound sense of scale, which is achieved by integrating the geometry into an atmospheric medium rather than a sterile vacuum.

Depth Fog and Exponential Height Fog

Depth fog provides atmospheric perspective by blending the geometry color with a background scattering color based on distance. A simple linear interpolation based on depth looks synthetic and flat. Instead, physically based rendering employs an exponential decay formula based on the Beer-Lambert law: [Figure omitted from source export], where [Figure omitted from source export] is the optical depth34.

To enhance vertical scale and ground floating abstract structures, exponential height fog is utilized. This algorithm computes fog density based on both the distance from the camera and the world-space height of the pixel36. This allows abstract forms to rise out of a dense, luminescent mist, providing a continuous gradient that roots floating geometry into a cohesive environment. For premium pipelines, this can be combined with ray-marched volumetric scattering, where light shafts are evaluated by stepping through the depth buffer toward the light source, heavily amplifying the atmospheric glow34.

Depth of Field and the Circle of Confusion

Depth of Field (DoF) simulates the focal plane of a physical camera, isolating specific geometry while smoothly obscuring the foreground and background. The diameter of the blur is dictated by the Circle of Confusion (CoC). Derived from thin-lens optics, the CoC is mathematically expressed as:

[Figure omitted from source export]

where [Figure omitted from source export] is the focal length, [Figure omitted from source export] is the aperture f-number, [Figure omitted from source export] is the focus distance, and [Figure omitted from source export] is the depth of the subject38.

To optimize this for real-time manipulation and artist intuition, the formula is often simplified to use an arbitrary aperture scalar, generating a signed CoC where negative values indicate foreground blur and positive values indicate background blur40:

[Figure omitted from source export]

Preserving the sign of the CoC is critical; it allows the shader to differentiate between foreground objects bleeding over the background and background objects being obscured by sharp foreground edges40. DoF is notoriously expensive to render because a true scatter approach—where each pixel distributes its energy to surrounding pixels based on its CoC—requires complex depth sorting and immense compute power41.

Consequently, high-end pipelines use a gather approach, blurring the image into multiple background and foreground buffers, then interpolating between them based on the per-pixel CoC43. To create aesthetically pleasing, cinematic blur, the gather phase utilizes disk sampling governed by the Golden Angle, generating a spiral pattern that prevents the banding and ringing artifacts common in standard grid-based sampling40. By mapping a bokeh shape texture to these sample offsets, artists can introduce hexagonal, anamorphic, or custom bokeh shapes that drastically elevate the perceptual richness of the out-of-focus areas40.

Temporal Dynamics, Anti-Aliasing, and Noise Injection

In static imagery, subtle gradients are prone to banding, and hard edges exhibit stair-stepping. In motion, aliasing creates a flickering, jagged mess that shatters the mesmerism of abstract graphics. Temporal accumulation and noise injection are required to stabilize the image.

Temporal Accumulation and Anti-Aliasing (TAA)

Temporal Anti-Aliasing (TAA) relies on sub-pixel camera jitter to generate multi-sample patterns over time without the immense cost of spatial supersampling44. The renderer shifts the projection matrix by microscopic, sub-pixel amounts dictated by a low-discrepancy Halton sequence, ensuring even distribution across the pixel area44.

During the temporal resolve phase, motion vectors track the screen-space movement of pixels between the previous and current frame. This velocity buffer allows the history buffer to be reprojected, finding where the specific surface existed on the screen one frame ago44. However, when objects move, they disocclude the background, revealing pixels that have no valid history. If the history buffer is blindly blended, "ghosting" or afterimages occur, manifesting as dark trails behind moving glowing geometry44.

To prevent this, the pipeline enforces strict neighborhood clamping or variance clipping44. The reprojected history pixel's color is clamped to an Axis-Aligned Bounding Box (AABB) defined by the minimum and maximum color values—or a variance bounding box computed via the mean and standard deviation—of the immediate [Figure omitted from source export] pixel neighborhood in the current frame44. If the history pixel falls outside this color space, it is rejected, maintaining temporal stability without ghosting45. Alternatively, afterimages can be artistically embraced in abstract graphics by intentionally loosening the variance clamping multiplier, leaving a deliberate, fading trail behind fast-moving forms.

Motion Blur and Tile-Max Velocity

Motion blur adds crucial fluidity to fast-moving procedural shapes, simulating the shutter speed of a physical camera. High-end implementations rely on the McGuire reconstruction filter, utilizing a tile-max velocity buffer46. Computing motion blur on a per-pixel basis often results in foreground objects failing to blur over static backgrounds. The tile-max approach divides the screen into tiles (e.g., [Figure omitted from source export] pixels) and calculates the maximum velocity vector for each tile47. This dilated velocity buffer ensures that fast-moving foreground objects correctly bleed their color over static background objects, creating a plausible, cinematic motion blur that preserves geometric continuity47.

Film Grain and Blue-Noise Dithering

Procedural gradients, especially in dark, atmospheric scenes heavily reliant on fog and bloom, suffer from severe quantization banding when compressed into standard 8-bit or 10-bit display formats49. Injecting noise breaks up these mathematical bands, distributing the quantization error across the spatial domain50.

Standard white noise is visually harsh and introduces unwanted low-frequency artifacts. Premium pipelines instead use blue-noise dithering52. Blue noise lacks low-frequency components, meaning it has no clumps or discernible patterns; it appears as a perfectly uniform, high-frequency static that the human visual system naturally filters out52. Applying blue-noise dithering just before the final quantization step preserves the perceptual richness of infinite-precision shader math while adhering to the hard limits of the display buffer, resulting in perfectly smooth, continuous gradients49.

Architectural Synthesis and Implementation

The sequence of operations in a post-processing pipeline is not arbitrary; it is strictly dictated by the mathematics of light and human perception.

Recommended Ordering of Rendering Passes

A physically and perceptually sound architecture executes in the following justified order:

1. Velocity and Depth Resolve: (Linear Space) Extracts motion vectors and depth buffers necessary for subsequent temporal passes.

2. Temporal Accumulation (TAA): (Linear Space) Reconstructs sub-pixel detail using jitter and history buffers. This must occur early; if deferred until after bloom or depth of field, the TAA will falsely accumulate the blurred pixels, resulting in severe smearing and loss of detail.

3. Depth of Field: (Linear Space) Focuses the geometry based on the CoC. Executing this in linear space ensures out-of-focus bright spots correctly expand into accurate, energy-conserving bokeh shapes rather than unnaturally dimming.

4. Motion Blur: (Linear Space) Uses the tile-max velocity buffer to smear the focused geometry according to simulated shutter speed.

5. Bloom Extraction and Dual-Filtering: (Linear Space) Extracts values above the soft-knee threshold and processes the multi-scale glow. Added back to the main buffer via fractional weights.

6. Exposure and Tone Mapping: (Transition from Linear to Perceptual Space) Transforms the unbounded HDR linear signal into the display's gamut using AgX, Uncharted 2, or Khronos Neutral.

7. Color Grading: (Perceptual Space) Applies 3D LUTs or ASC CDL to stylize the compressed dynamic range.

8. Optical Distortions: (Perceptual Space) Applies chromatic aberration, barrel distortion, and vignetting. These simulate camera lens artifacts which alter the recorded, tone-mapped image, hence they are applied post-tonemap.

9. Blue-Noise Dithering and UI Overlay: (Perceptual Space) Dithers the final image to prevent banding before final quantization and presentation to the swapchain.

Adaptive Quality Tiers

To maintain consistent framerates without fundamentally altering the artistic composition, the pipeline must scale gracefully across hardware capabilities.

Enterprise Cost Matrix

The deployment of a premium post-processing stack requires strict budgeting of GPU compute time and memory bandwidth. The following cost matrix outlines the relative expense of each node.

Artifact Identification Checklist

When developing abstract procedural graphics, identifying the root mathematical cause of visual degradation is critical for a technical artist.

Fifteen Visual Recipes for Dreamlike Graphics

To harness the comprehensive architecture described, technical artists combine parameters to create distinct, highly stylized visual profiles. The following fifteen recipes define precise parameter configurations to achieve mesmerizing procedural aesthetics without overwhelming the geometric foundation.

The synthesis of these advanced post-processing techniques—anchored by physically correct linear mathematics, bounded by intelligent temporal accumulation, and finalized through perception-aware color science—ensures that procedural and abstract geometry achieves its maximum artistic potential. By rigorously controlling how light bleeds, how lenses distort, and how colors are mapped to physical hardware, developers can create truly mesmerizing real-time experiences.

Works cited

1. glsl-tone-map/agx.glsl at main \- GitHub, https://github.com/dmnsgn/glsl-tone-map/blob/main/agx.glsl

2. WebGPU \- W3C, https://www.w3.org/TR/webgpu/

3. GPUCanvasContext: configure() method \- Web APIs | MDN, https://developer.mozilla.org/en-US/docs/Web/API/GPUCanvasContext/configure

4. SurfaceColorSpace in wgpu \- Rust \- Docs.rs, https://docs.rs/wgpu/latest/x86\_64-apple-darwin/wgpu/enum.SurfaceColorSpace.html

5. webgpu-hdr/EXPLAINER.md at main \- GitHub, https://github.com/ccameron-chromium/webgpu-hdr/blob/main/EXPLAINER.md

6. Artist-Friendly HDR With Exposure Values \- Nathan Reed, https://www.reedbeta.com/blog/artist-friendly-hdr-with-exposure-values/

7. Dynamic Range, Exposure And Tonemapping \- Seena Burns, https://seenaburns.com/dynamic-range/

8. FFmpeg Filters Overview and Examples | PDF \- Scribd, https://www.scribd.com/document/693828187/ffmpeg-filters

9. ffprobe Documentation \- Index of /docs, https://gensoft.pasteur.fr/docs/ffmpeg/4.3.1/ffprobe-all.html

10. ffplay-all: FFplay media player | Man Page | Commands | ffmpeg, https://www.mankier.com/1/ffplay-all

11. Options \- libplacebo, https://libplacebo.org/options/

12. mpv(1) \- FreeBSD 15.1-RELEASE and Ports.quarterly Manual Pages, https://man.freebsd.org/cgi/man.cgi?mpv(1)

13. Lypheo/vs-placebo: libplacebo-based debanding, scaling ... \- GitHub, https://github.com/Lypheo/vs-placebo

14. A Closer Look At Tone Mapping \- The Danger Zone, https://therealmjp.github.io/posts/a-closer-look-at-tone-mapping/

15. Blog \- Stefan Groenewoud, https://stefangroenewoud.artstation.com/blog?page=4

16. Tonemapped Natural Color Script \- Sentinel Hub custom scripts, https://custom-scripts.sentinel-hub.com/custom-scripts/sentinel-2/tonemapped\_natural\_color/

17. Filmic Tonemapping with Piecewise Power Curves, https://filmicworlds.com/blog/filmic-tonemapping-with-piecewise-power-curves/

18. Created and applied LUT in VFB looks different \- Chaos Forum, https://forums.chaos.com/t/created-and-applied-lut-in-vfb-looks-different/113563

19. Support AgX tone mapping \#27362 \- mrdoob/three.js \- GitHub, https://github.com/mrdoob/three.js/issues/27362

20. PBR Neutral Tone Mapping \-, https://modelviewer.dev/examples/tone-mapping

21. Physically Based Tone Mapping and Perceptual Fidelity in Gran, https://blog.selfshadow.com/publications/s2025-shading-course/pdi/s2025\_pbs\_pdi\_slides\_v1.1.pdf

22. Khronos PBR Neutral Tone Mapper Released for True-to-Life Color, https://www.khronos.org/news/press/khronos-pbr-neutral-tone-mapper-released-for-true-to-life-color-rendering-of-3d-products

23. Bandwidth-efficient graphics \- Arm Community, https://community.arm.com/cfs-file/\_\_key/communityserver-blogs-components-weblogfiles/00-00-00-20-66/siggraph2015\_2D00\_mmg\_2D00\_marius\_2D00\_notes.pdf

24. Screen Post Processing Effects Chapter 4: Kawase Blur and Its, https://medium.com/@uwa4d/screen-post-processing-effects-chapter-4-kawase-blur-and-its-implementation-5b0ce98b8706

25. Bandwidth-Efficient Rendering \- Froyok, https://www.froyok.fr/blog/2024-01-breakdown-syndicate/resources/presentations/siggraph2015-mmg-marius-slides.pdf

26. Bloom \- Catlike Coding, https://catlikecoding.com/unity/tutorials/advanced-rendering/bloom/

27. What is Chromatic Aberration in Gaming? \- Spotlight FX, https://spotlightfx.com/blog/what-is-chromatic-aberration-in-games

28. What effect does Chromatic Abberation have on our perception of, https://www.reddit.com/r/explainlikeimfive/comments/7lqmep/eli5\_what\_effect\_does\_chromatic\_abberation\_have/

29. Rendering Chromatic Eye Aberration Improves Accommodation and, https://www.researchgate.net/publication/317681552\_ChromaBlur\_Rendering\_Chromatic\_Eye\_Aberration\_Improves\_Accommodation\_and\_Realism\_in\_HMDs

30. ChromaBlur: Rendering Chromatic Eye Aberration Improves, https://www.rose-hulman.edu/class/cs/csse451/Research%20paper/2019-papers/ChromaBlur.pdf

31. Vergence–accommodation conflict \- Wikipedia, https://en.wikipedia.org/wiki/Vergence%E2%80%93accommodation\_conflict

32. What is Chromatic Aberration in Games | Boris FX, https://borisfx.com/blog/what-is-chromatic-aberration-in-games/

33. Control of accommodation by longitudinal chromatic aberration and, https://jov.arvojournals.org/article.aspx?articleid=2191954

34. Volumetric Fog | Mattias Strand, https://www.mattiasstrand.com/posts/volumetric-fog/

35. Raindrops on a puddle \- Nikos Papadopoulos, https://www.4rknova.com/blog/2026/02/08/raindrops-on-puddle

36. Volumetric Fog Rendering \- DSpace, https://dspace.ut.ee/server/api/core/bitstreams/006047ad-04c7-4bc7-8151-68b1a6954040/content

37. Physically Based and Unified Volumetric Rendering in Frostbite, https://www.slideshare.net/slideshow/physically-based-and-unified-volumetric-rendering-in-frostbite/51840934

38. Deriving blur from real optical formulae, https://computergraphics.stackexchange.com/questions/7873/deriving-blur-from-real-optical-formulae

39. DEPTH OF FIELD IMPLEMENTATION WITH OPENGL, https://pages.mtu.edu/\~shene/PUBLICATIONS/2004/CCSC-MW-2004-depth\_of\_field.pdf

40. Building a Custom Miniature-Style DoF in UE with HLSL, https://dev.to/taku25/unpacking-the-math-building-a-custom-miniature-style-dof-in-ue-with-hlsl-2118

41. Depth of Field \- Research Unit of Computer Graphics | TU Wien, https://www.cg.tuwien.ac.at/research/publications/2015/Kollmann-2015-DoF/Kollmann-2015-DoF-thesis.pdf

42. Depth-of-Field Rendering with Multiview Synthesis, https://publications.graphics.tudelft.nl/rails/active\_storage/blobs/redirect/eyJfcmFpbHMiOnsibWVzc2FnZSI6IkJBaHBBaHNLIiwiZXhwIjpudWxsLCJwdXIiOiJibG9iX2lkIn19--30bf1d37b71134097d077e1194daf03adf1e0dae/LeeDOF.pdf

43. Chapter 28\. Practical Post-Process Depth of Field \- NVIDIA Developer, https://developer.nvidia.com/gpugems/gpugems3/part-iv-image-effects/chapter-28-practical-post-process-depth-field

44. DirectX 12: TAA \- CosmicLearn, https://www.cosmiclearn.com/dx12/taa.php

45. Unreal Engine5 Customization with SSGI and Optimization based, https://github.com/whlee503/ue5\_custom\_ssgi

46. SMAA 1TX vs 2TX in Real-Time Rendering \- Scribd, https://www.scribd.com/document/220450776/Sousa-Graphics-Gems-CryENGINE3

47. Next-Gen Post Processing in CoD: AW | PDF \- Scribd, https://www.scribd.com/presentation/363243286/Next-Generation-Post-Processing-in-Call-of-Duty-Advanced-Warfare-v18-pptx

48. Graphics Gems from CryENGINE 3 (Siggraph 2013\) | PDF \- Slideshare, https://www.slideshare.net/slideshow/graphics-gems-from-cryengine-3-siggraph-2013/25052583

49. Correct sRGB Dithering | The Tenth Planet, http://www.thetenthplanet.de/archives/5367

50. Tech Note: Shader Snippets for Efficient 2D Dithering, https://developers.meta.com/horizon/blog/tech-note-shader-snippets-for-efficient-2d-dithering/

51. Rendering in Real Time with Spatiotemporal Blue Noise Textures, https://developer.nvidia.com/blog/rendering-in-real-time-with-spatiotemporal-blue-noise-textures-part-2/

52. Stable Fiddusion \- Acko.net, https://acko.net/blog/stable-fiddusion/

53. Free blue noise textures \- Moments in Graphics, https://momentsingraphics.de/BlueNoise.html

54. Dizzy Dithering \- Liam Appelbe, https://liamappelbe.medium.com/dizzy-dithering-2ae76dbceba1

55. The Art of Dithering and Retro Shading for the Web, https://blog.maximeheckel.com/posts/the-art-of-dithering-and-retro-shading-web/