DropBG

Local-first, privacy-friendly background remover for macOS.
No uploads, no subscriptions, no resolution limits.

Why DropBG?

• Local-first — images never leave your machine by default
• Free & unlimited — no per-image fees or resolution caps
• Fast on Apple Silicon — leverages CoreML / Neural Engine
• One-time setup — download the AI model once, then everything runs offline
• Optional cloud API — use Replicate, fal.ai, or remove.bg for cloud GPU processing

Features

• Drag-and-drop single or multiple images
• 6 AI models to choose from (BiRefNet Lite/Full, BEN2, RMBG 2.0, InSPyReNet, MODNet)
• Cloud API support — Replicate, fal.ai, remove.bg (bring your own key)
• Batch processing with per-image progress and auto-naming
• Background replacement — solid colors, gradients, or custom images
• AI upscaling — 2x/4x super-resolution via Real-ESRGAN
• Auto-crop — trim transparent edges automatically
• Before/after preview — press Space to toggle
• Configurable — model location, save folder, model switching without restart

Available Models

Local Models (runs 100% offline)

Model	Size	Tier	Best For
BiRefNet Lite	~200 MB	Default	Fast, good for most images
BiRefNet Full	~900 MB	Best quality	Complex backgrounds, high detail
BEN2	~219 MB	Best edges	Hair, fur & fine edges
RMBG 2.0	~514 MB	Best product	Clean ecommerce/product shots (manual download)
InSPyReNet	~300 MB	Best fuzzy	Fuzzy edges, hair strands & fine detail
MODNet	~13 MB	Lightweight	Real-time / portraits (legacy)

Cloud API Providers (optional, bring your own key)

Provider	Cost	Model	Notes
Replicate	~$0.0004/img	BiRefNet on A100 GPU	Cheapest, same models on cloud GPUs
fal.ai	~$0.018/img	BiRefNet + RMBG 2.0	Fast & reliable
remove.bg	~$0.10/img	Proprietary	Best polish, most mature API

Tech Stack

Layer	Choice	Why
App	Tauri 2 (Rust + React)	Lightweight, native feel
AI Inference	ort (ONNX Runtime) + CoreML EP	Apple Neural Engine acceleration
AI Upscale	Real-ESRGAN x4plus (ONNX)	High-quality super-resolution
Image	image crate + ndarray	PNG read/write with alpha channel
Frontend	React 19 + TypeScript	Fast, component-based UI

Getting Started

Prerequisites: Rust toolchain, Bun

# clone
git clone https://github.com/whereissam/DropBG && cd DropBG

# install frontend dependencies
bun install

# run in dev mode
cargo tauri dev

On first launch, DropBG asks you to download an AI model (~200 MB). After that, everything runs 100% offline.

Installing from DMG (no code signing)

Since DropBG doesn't have an Apple Developer certificate yet, macOS will block the app on first launch:

1. Try to open DropBG — macOS shows a warning
2. Go to System Settings → Privacy & Security
3. Click Open Anyway next to the DropBG message
4. Or run: xattr -cr /Applications/DropBG.app

See docs/USAGE.md for detailed instructions.

Build for Release

cargo tauri build

The DMG/app bundle will be in src-tauri/target/release/bundle/.

Documentation

• Usage Guide — installation, features, models, troubleshooting
• TODO / Roadmap — development phases and progress

Project Structure

DropBG/
├── src-tauri/                # Rust backend
│   ├── src/
│   │   ├── lib.rs            # Tauri entry + command registration
│   │   ├── commands.rs       # IPC command handlers
│   │   ├── inference/        # ONNX Runtime session, pre/post processing, upscale
│   │   ├── imaging/          # Auto-crop, background replacement
│   │   └── model/            # Model downloader + config management
│   └── Cargo.toml
├── src/                      # React frontend
│   ├── App.tsx               # Main app (stage-based routing)
│   ├── tauri.ts              # Typed Tauri invoke wrappers
│   └── components/           # UI components
├── web/                      # Landing page (Astro)
├── docs/
│   ├── USAGE.md              # Usage guide
│   └── TODO.md               # Roadmap
└── README.md

License

MIT