Real-Time Editor Tools for Unity

VHS 1980 VR Effects

The VHS 1980 VR Effects package provides authentic 1980s VHS tape aesthetic effects for Unity projects. All effects are designed to work seamlessly in both standard and VR environments, recreating the nostalgic look of vintage video recordings.

Effect Categories

The system is organized into 7 main categories:

• General Settings — Master intensity, resolution control, motion blur
• Chroma Processing — Color bleed, wobble, offset, edge enhancement, film grain
• Scanline Effects — CRT scanlines, VR-compatible tape lines with distortion
• Signal Interference — Head switching noise, static bars, signal dropout, rainbow banding
• CRT Display Effects — Ghosting, vignette, phosphor glow, barrel distortion
• Tape Playback Effects — Tracking issues, artifacts, wobble, pause jitter
• Color Grading — Film response, saturation, temperature, contrast/exposure

1980s Movie Presets

5 authentic movie-style presets recreate iconic visual aesthetics:

Preset Overview

VR Optimization

Special attention has been given to VR comfort. Tape Lines are specifically designed for VR environments, while effects that may cause discomfort (Barrel Distortion, Bottom Roll) can be easily disabled. All effects work correctly in both eyes independently.

Requirements

• Unity 6.x LTS or later
• Compatible with Unity URP for VR
• VR Ready (OpenXR, Oculus, SteamVR)

Gaussian Splat Editor for Unity

A powerful Unity Editor plugin designed for editing and manipulating 3D Gaussian Splats directly within the Unity environment. Streamline your workflow with intuitive visual tools for splat selection, transformation, and real-time preview.

Key Features

• Visual Selection Tools — Select individual splats or groups using familiar Unity-style selection methods
• Transform Operations — Move, rotate, and scale splats with precision gizmos
• Real-Time Preview — See changes instantly in the editor viewport
• Import/Export PLY — Full support for PLY file format, the standard for Gaussian Splat data
• Batch Operations — Apply transformations to thousands of splats simultaneously
• VR-Ready — Edit splats in VR mode for immersive manipulation
• Undo/Redo — Full history support with Unity's native undo system

Workflow Integration

The editor integrates seamlessly with Unity's existing tools and workflows. Use the familiar Inspector panel to adjust splat properties, leverage Unity's scene view for navigation, and take advantage of the built-in undo system for non-destructive editing.

Whether you're cleaning up captured scenes, compositing multiple splat clouds, or preparing assets for real-time applications, this plugin provides the tools you need without leaving Unity.

Requirements

• Unity 2021.3 LTS or later
• Compatible with URP, HDRP, and Built-in Render Pipeline
• Windows, macOS, and Linux editor support

Real-Time Fake Relighting for 3D Gaussian Splatting

A Novel Approach to Dynamic Lighting in Billboard-Based Volumetric Rendering

Abstract

I present a novel approach for real-time fake relighting of 3D Gaussian Splats that approximates physically-based lighting without requiring explicit surface normals or extensive preprocessing. My method introduces custom attenuation models for directional, point, and spot lights that work specifically with the billboard-based rendering nature of Gaussian Splatting.

By treating view direction as implicit surface orientation and applying carefully tuned distance-based and angular attenuation, I achieve plausible lighting effects at minimal computational cost suitable for real-time applications including VR. The system maintains 60+ FPS performance while supporting multiple dynamic lights, shadows, and SSAO integration.

Keywords: Gaussian Splatting · Real-time Rendering · Dynamic Lighting · Attenuation Models · Volumetric Rendering

1. Introduction

1.1 Background

3D Gaussian Splatting [Kerbl et al. 2023] represents scenes as collections of anisotropic 3D Gaussians with spherical harmonic (SH) appearance encoding. While SH coefficients capture view-dependent appearance from the original capture conditions, they lack the ability to respond to dynamic lighting changes—a critical limitation for interactive applications.

Traditional surface-based relighting relies on explicit surface normals for Lambert/Phong shading, material properties (albedo, roughness, metallicity), and BRDF evaluation at each surface point. Gaussian Splats present unique challenges: no explicit surface representation, billboard quads oriented toward camera, view-dependent SH color encoding, and alpha-blended volumetric composition.

1.2 Contributions

I introduce a fake relighting system that provides:

1. View-as-Normal Approximation: Uses view direction as implicit surface orientation
2. Custom Attenuation Models: Squared distance falloff for point lights with smooth spatial transitions
3. Directional Light Enhancement: Combines shadow mapping with custom distance and angular attenuation
4. Spot Light Cone Filtering: Smooth inner/outer cone transitions with distance falloff
5. GPU-Accelerated Implementation: Efficient compute shader-based light buffer management

2. Billboard Primitives in Gaussian Splatting

2.1 Understanding Billboard Rendering

Gaussian Splatting renders each splat as a billboard primitive—a quad that always faces the camera. This technique, commonly used in particle systems and volumetric effects, presents unique challenges for dynamic lighting since traditional surface-based shading assumes fixed surface orientations.

Unlike polygonal meshes where surface normals are defined by geometry, billboard primitives dynamically reorient themselves each frame to face the viewer. This view-dependent orientation is fundamental to the Gaussian Splatting representation, where each 3D Gaussian is projected into screen space as an anisotropic 2D footprint.

Figure 1: Billboard primitives maintain camera-facing orientation regardless of viewpoint

2.2 Lighting Challenges for Billboards

Traditional lighting models (Lambert, Phong, BRDF) require stable surface normals computed from geometry. For billboards, three approaches exist:

1. Ignore lighting: Treat billboards as unlit sprites (common in legacy particle systems)
2. Fixed normals: Assign arbitrary normals independent of view direction [Unity VFX Graph 6-way lighting approach]
3. View-as-normal: Use the view direction itself as the effective normal (my approach)

2.3 View-as-Normal Justification

My approach treats the view direction as the effective surface normal. This is justified for Gaussian Splats because:

• Capture consistency: Splats represent surfaces that were visible during capture—by definition, facing toward the camera
• Billboard nature: Since splats always orient toward the viewer, the view direction approximates their implicit surface orientation
• Zero overhead: View direction is already computed for billboard orientation, requiring no additional data
• Perceptual validity: Human perception of lighting on camera-facing surfaces is less sensitive to normal inaccuracies than on oblique surfaces

3. Mathematical Foundation

3.1 View-Direction as Surface Normal

Traditional surface rendering computes lighting using surface normal n. For Gaussian Splats, I lack explicit normals. My key insight: the view direction inherently encodes surface orientation for billboard primitives.

3.2 Directional Light Model

For directional lights (e.g., sun), the lighting computation incorporates shadow attenuation, distance-based falloff, and angular atmospheric effects:

3.3 Point Light Model

Point lights require distance-based attenuation. I use a squared falloff model that provides smooth transitions while maintaining exact zero at the light boundary:

Mathematical Properties:

• Smooth falloff: The squared term creates smooth transitions
• Zero at boundary: Exactly 0 at range R
• No division: Avoids numerical instability
• Non-linear: More aggressive falloff than linear

3.4 Spot Light Model

Spot lights extend point lights with angular constraints, providing directional control through cone attenuation:

3.5 Combined Lighting Model

The final lighting applied to each Gaussian Splat integrates ambient, directional, point, and spot contributions:

4. Algorithmic Framework

4.1 Rendering Pipeline

The rendering pipeline follows a multi-stage evaluation: (1) Vertex Stage transforms splat positions to world space and computes view-direction normals; (2) Fragment Stage evaluates Gaussian alpha, samples ambient occlusion, and computes lighting; (3) Composition blends lit splats using alpha-weighted accumulation.

4.2 Light Accumulation Algorithm

The lighting computation follows an additive accumulation strategy with computational complexity O(1 + N_point + N_spot) per fragment.

5. Performance Analysis

Per-fragment instruction count: Base(30) + 12×N_point + 18×N_spot instructions

Table 1: Per-Fragment Cost Breakdown

Table 2: Performance Measurements (RTX 3080, 2560×1440, 2M splats)

Table 3: Light Count Scaling

6. Visual Results and Validation

My method produces:

• Smooth Light Transitions: Squared attenuation avoids harsh boundaries
• Plausible Shadows: Integration with shadow mapping maintains consistency
• View-Dependent Shading: Splats facing away from lights correctly darken
• Ambient Preservation: SH ambient prevents over-darkening

6.1 Limitations

1. No Specular Response: View-as-normal approximation prevents proper specular highlights
2. Limited Normal Detail: Cannot represent surface microstructure
3. Translucency Issues: Alpha blending complicates multi-scattering effects
4. Shadow Aliasing: Inherited from shadow map resolution

7. Conclusion

I have presented a practical fake relighting system for 3D Gaussian Splatting that achieves plausible dynamic lighting at real-time performance. The key contributions are: (1) View-as-Normal Approximation that leverages billboard orientation for implicit shading; (2) Optimized Attenuation with squared falloff that balances quality and performance; (3) Production-Ready Implementation that is stable, efficient, and artist-friendly.

The system supports complex scenes with multiple dynamic lights while maintaining 60+ FPS, making it suitable for interactive applications, VR experiences, and real-time visualization.

References