◈ NeoSketch

Offline AI Art Studio. No Cloud. No Compromise.

NeoSketch is a lightweight, fully offline AI-powered artistic photo editor built for Android and Windows.
Flutter frontend. Rust engine. Python AI tooling. Zero cloud dependency.
Professional artistic transformations — running entirely on your device.

---

◈ Why NeoSketch?

Most AI photo editors push everything to the cloud — your images, your data, your privacy. NeoSketch doesn't.

Principle	What it means
🔒 Offline-First	All AI processing runs on-device. No internet required
🧠 Privacy-Focused	Your images never leave your phone or PC
⚡ Rust-Powered Engine	High-performance backend built in Rust for raw speed
🎨 Hyper-Realistic Art	Real artistic rendering, not cheap Instagram filters
🪶 Lightweight	Optimized for low RAM, small APK, fast launch
🖥️ Cross-Platform	Android + Windows from a single codebase
🔧 Modular Architecture	Every component is pluggable and extendable
🤖 AI-Enhanced	ONNX + TFLite models for real on-device AI inference

---

◈ Features

🖼️ Core Features

• Offline AI artistic photo editor
• Real-time preview rendering
• GPU-aware optimization
• Project autosave system
• Preset management system
• Unlimited undo / redo
• Layer editing support
• Drag-and-drop image import
• Batch export
• Keyboard shortcuts (Windows)
• Touch gesture support (Android)
• Responsive adaptive UI

🎨 Artistic Styles

Pencil Sketch    │  Charcoal Sketch  │  Ink Art
Watercolor       │  Oil Painting     │  Anime
Manga            │  Comic            │  Cyberpunk
Pixel Art        │  Pop Art          │  Vintage
Clay Style       │  Neon Art         │  AI Stylization
+ more advanced styles in roadmap

🤖 AI Features

• AI Upscale & Super Resolution
• AI Detail Enhancement
• AI Portrait Enhancement
• Background Removal
• Object Removal
• AI Colorization
• AI Denoising
• AI Style Transfer
• Uncrop / Outpainting
• AI Texture Enhancement

🛠️ Editing Tools

Crop  │  Resize  │  Rotate  │  Flip  │  Perspective Correction
Brightness  │  Contrast  │  Saturation  │  Blur  │  Sharpen
Brush Controls  │  Gradient Controls  │  Texture Overlays

---

◈ Tech Stack

Architecture Overview

┌─────────────────────────────────────────────────────────┐
│                    Flutter UI (Dart)                     │
│              Riverpod State │ GoRouter Navigation        │
└────────────────────┬────────────────────────────────────┘
                     │ flutter_rust_bridge (FFI)
┌────────────────────▼────────────────────────────────────┐
│                   Rust Engine                            │
│         rayon │ OpenCV-Rust │ image-rs │ FFI             │
└────────────────────┬────────────────────────────────────┘
                     │ Model Inference
┌────────────────────▼────────────────────────────────────┐
│             ONNX Runtime │ TensorFlow Lite               │
│                  On-Device AI Models                     │
└────────────────────┬────────────────────────────────────┘
                     │ Refined Output
┌────────────────────▼────────────────────────────────────┐
│             Rust Post-Processing & Refinement            │
└────────────────────┬────────────────────────────────────┘
                     │
┌────────────────────▼────────────────────────────────────┐
│               Flutter Preview & Export                   │
└─────────────────────────────────────────────────────────┘

Stack Breakdown

Layer	Technology
Frontend	Flutter, Dart, Riverpod, GoRouter
Backend Engine	Rust, flutter_rust_bridge, FFI, rayon, OpenCV-Rust, image-rs
AI Runtime	ONNX Runtime, TensorFlow Lite
Python AI Tooling	PyTorch, OpenCV, Diffusers, Pillow, NumPy, TensorFlow, Scikit-image
Storage	Isar Database, Local project storage, Cached thumbnails, Model storage

---

◈ Project Structure

NeoSketch/
│
├── lib/                        # Flutter frontend
│   ├── core/                   # App config, themes, constants
│   ├── features/               # Feature modules (editor, ai_lab, export)
│   ├── shared/                 # Shared widgets and utilities
│   └── main.dart
│
├── rust/                       # Rust engine
│   ├── src/
│   │   ├── engine/             # Core rendering engine
│   │   ├── filters/            # Artistic filter implementations
│   │   ├── ai/                 # ONNX/TFLite inference bridge
│   │   └── lib.rs
│   └── Cargo.toml
│
├── python/                     # Python AI tooling
│   ├── model_export/           # Export PyTorch → ONNX/TFLite
│   ├── training/               # Model training scripts
│   ├── preprocessing/          # Dataset prep utilities
│   └── requirements.txt
│
├── assets/
│   ├── models/                 # Bundled ONNX / TFLite models
│   ├── presets/                # Default artistic presets
│   └── icons/
│
└── shared/                     # Shared utilities and configs

---

◈ Performance Goals

Target	Goal
UI Frame Rate	60 FPS stable
RAM Usage	Low, optimized per device class
Rendering	Fully async, non-blocking
GPU	GPU-aware optimization where available
Processing	100% offline
APK Size	Lightweight, minimal bloat
Rendering Threads	Multithreaded via Rust rayon
Model Loading	Efficient cached inference

---

◈ Screenshots

Screenshots will be added after first stable build

Screen	Preview
Home Dashboard	[placeholder]
Editor View	[placeholder]
AI Lab	[placeholder]
Sketch Mode	[placeholder]
Watercolor Mode	[placeholder]
Anime Mode	[placeholder]
Export Screen	[placeholder]
Desktop Workspace	[placeholder]

---

◈ Installation & Setup

Prerequisites

• Flutter SDK >=3.0.0
• Rust >=1.75 (stable)
• Python >=3.10
• Android SDK / NDK (for Android build)
• ONNX Runtime

---

1. Clone the Repo

git clone https://github.com/Tcode-Motion/NeoSketch.git
cd NeoSketch

2. Flutter Setup

flutter pub get

3. Rust Engine Setup

cd rust
cargo build --release
cd ..

4. Python AI Tooling Setup

cd python
pip install -r requirements.txt
cd ..

5. ONNX Runtime Setup

# Download and place models in assets/models/
# See python/model_export/ for export scripts
python python/model_export/export_all.py

6. Run — Android

flutter run -d android

7. Run — Windows

flutter run -d windows

---

◈ Roadmap

Phase	Feature	Status
v1.0	Core editor + 10 artistic styles	🔨 In Progress
v1.1	AI Upscale + Background Removal	📋 Planned
v1.2	Batch export + Preset system	📋 Planned
v2.0	AI Video Filters	🔮 Future
v2.0	Live Camera Effects	🔮 Future
v2.1	Plugin Marketplace	🔮 Future
v2.2	Community Presets	🔮 Future
v2.3	Custom AI Model Downloads	🔮 Future
v3.0	macOS Support	🔮 Future
v3.0	Linux Support	🔮 Future

---

◈ Contributing

Pull requests are welcome. If you want to contribute:

• 🧩 Modular contributions — each filter, AI feature, and UI screen is isolated
• 🤖 AI model optimization — help reduce model size and improve inference speed
• 🎨 UI improvements — new artistic styles, better previews, new presets
• ⚡ Performance — Rust engine optimizations, async improvements, GPU utilization
• 🐛 Bug fixes — open an issue first, then PR

# Fork → Clone → Branch → PR
git checkout -b feature/your-feature-name

Please keep PRs focused. One feature per PR.

---

◈ Design Philosophy

NeoSketch is built around a few non-negotiable principles:

• Real artistic rendering — no fake Instagram-style filters. Every style is designed to look like actual art
• Human-like artistic quality — the goal is output that looks hand-made, not AI-generated
• Offline privacy-first — your images, your device, your data
• Modular architecture — every feature is independently swappable
• Native performance — Rust where it matters, Flutter where it shines
• Clean scalable codebase — built to grow without breaking

---

◈ License

MIT License — see LICENSE for full text

---

◈ Credits

Built on the shoulders of giants:

• Flutter — cross-platform UI framework
• Rust — systems-level performance engine
• ONNX Runtime — on-device AI inference
• TensorFlow Lite — mobile AI runtime
• OpenCV — computer vision backbone
• PyTorch — model training
• Diffusers — diffusion model tooling
• The entire open-source AI community

---

NeoSketch aims to become a professional offline AI creative studio —
combining the best ideas from Procreate, Photoshop, Prisma, Lightroom,
and Stable Diffusion workflows into a single lightweight cross-platform experience.

Built solo. Shipped with intention.

Tcode-Motion

◈ Dark by default. Offline by design. Powerful by choice.