A Unity-based mobile game featuring ink-based life simulation with hierarchical rendering and machine learning stylization.

Inkling is a revolutionary mobile game where players interact with a living world of ink-based organisms. The game features advanced fluid simulation, cellular automata, and machine learning-driven stylization to create emergent gameplay experiences.

• Hierarchical Simulation Layers (HSL): Multi-scale simulation from molecular to organism level
• ML Stylization Pipeline: Real-time inference using trained U-Net models for artistic rendering
• Foveated Rendering: High-detail central zone with optimized peripheral rendering
• Physics Integration: Hybrid approach using Unity DOTS and classical physics

Inkling/
├── Assets/
│   └── Inkling/
│       ├── Runtime/
│       │   ├── Systems/
│       │   │   ├── SimulationLOD0/    # Core molecular simulation
│       │   │   ├── Inference/         # ML inference pipeline
│       │   │   └── Foveation/         # Foveated rendering system
│       │   └── Dev/
│       │       ├── DevOverlay/        # Debug visualization
│       │       └── RecordReplay/      # Deterministic replay system
│       ├── Shaders/                   # Compute and rendering shaders
│       └── Editor/                    # Unity Editor extensions
├── ProjectSettings/                   # Unity configuration
└── Packages/                          # Dependencies via MagiUnityDependencyManager

• Unity 2022.3 LTS or newer
• NVIDIA GPU with compute shader support (for development)
• Git LFS for binary assets

1. Ensure your Unity project root is ./Inkling (nested)
2. Run dependency manager to set up packages:

   ../MagiUnityDependencyManager/magi-deps.ps1 apply -ProjectPath ./Inkling -Strict

3. Open the Unity project at ./Inkling in Unity Editor
4. Import required packages via Unity Package Manager:
   • Unity Sentis (ML inference)
   • Universal Render Pipeline (URP)
   • Unity DOTS packages

• Core simulation logic in Assets/Inkling/Runtime/Systems/SimulationLOD0/
• Compute shaders in Assets/Inkling/Shaders/Simulation/
• Use deterministic fixed timestep for reproducibility

• ONNX models stored in Assets/Inkling/ML/Models/
• Inference pipeline in Assets/Inkling/Runtime/Systems/Inference/
• Dataset capture tools in Assets/Inkling/Editor/DatasetCapture/

• Simulation: ≤ 4-6ms per frame
• ML Inference: ≤ 3-5ms per frame
• Compositing: ≤ 2ms per frame
• Target Devices: iPhone 12, Pixel 6 (baseline)

• InkTools: Core fluid simulation and cellular automata systems
• InkModel: ML pipeline for training stylization models
• MagiUnityTools: Common Unity patterns and utilities
• MagiUnityDependencyManager: Package dependency management

• iOS: Xcode 14+ required
• Android: Minimum API Level 24 (Android 7.0)

• Tier 0: Shader-only stylization (no ML)
• Tier 1: Hybrid ML + shader rendering
• Tier 2: Full ML chain with foveation

See the magi-knowledge-repo-2 for comprehensive documentation including:

• Technical specification
• Implementation overview
• ML pipeline setup
• Art and design guidelines