Remove image backgrounds with local AI. Unlimited images, full resolution, zero uploads.
Pay once. Own it forever. No subscription.
remove.bg charges $9/month for 40 credits — and every photo you upload goes to their servers. BG Remover runs the same class of AI model on your own machine: no credits, no upload, no monthly bill, no resolution cap.
- Drag & drop — single images or whole batches (JPG / PNG / WebP)
- Before / after preview — checkerboard view of the transparent result, per image
- Flexible output — transparent PNG, white/black, or any custom background color
- Batch export — pick a folder, hit one button, watch the progress bar
- Full resolution — your 6000×4000 photo stays 6000×4000 (remove.bg free tier caps at 0.25 MP)
- 100% private — the AI model runs locally; your images never leave your computer
- Memory-safe queue — images process sequentially in an isolated worker process
The AI model (~80 MB) downloads automatically on the first removal — one time, clearly shown in the UI. After that, everything works fully offline.
Don't want to install Node and build from source? Grab the packaged Windows installer — one download, one click, done:
→ Get BG Remover on Whop — $24, one time, yours forever.
git clone https://github.com/bensblueprints/bg-remover.git
cd bg-remover
npm i
npm startRequires Node 18+ (tested on Node 24) and Windows / macOS / Linux.
| BG Remover | remove.bg | |
|---|---|---|
| Price | $24 once | $9/mo (40 credits) — $108/yr |
| Images | Unlimited | 40/month, then buy credits |
| Resolution | Full, always | Full res costs 1 credit per image |
| Privacy | 100% local — nothing uploaded | Every image uploaded to their servers |
| Offline | Yes (after one-time model download) | No |
| Batch processing | Yes, built in | Paid API / desktop app w/ credits |
| Source code | Open (MIT) | Closed |
$24 pays for itself in under 3 months — then it's free forever.
BG Remover uses @imgly/background-removal-node, an ONNX segmentation model (U²-Net family) running via onnxruntime-node — the same local-inference approach behind modern in-browser editors, but at desktop speed with no canvas limits.
- Electron main process owns a sequential job queue (one image at a time = predictable memory).
- Utility worker process runs inference, so the UI never freezes — even on a 40-image batch.
- sharp handles compositing flat-color backgrounds and preserving full input resolution.
No telemetry. No analytics. The only network call the app ever makes is the one-time model download.
- Electron — main + preload + renderer, utilityProcess worker
- @imgly/background-removal-node — local ONNX background removal
- sharp — image compositing / validation
- Plain HTML/CSS/JS renderer — dark mode, zero framework overhead
npm test # smoke test: real removal on a generated fixture
npx electron test/smoke-electron.js # same pipeline inside Electron's runtime
npm run dist # build Windows NSIS installer (electron-builder)The smoke test generates a fixture with sharp, runs the actual ONNX pipeline, and asserts the output PNG has a working alpha channel (transparent background corners, opaque subject).
MIT © 2026 Ben (bensblueprints)