Drop a low-res image. PixelLift reconstructs edges, recovers texture, and re-renders it at 8× to 12× the original — all in your browser.
PNG · JPG · WebP. Everything runs locally — your image never leaves the browser.
Multi-pass bicubic with Lanczos-style edge detection keeps lines crisp at extreme zoom.
Adaptive unsharp masking re-emphasizes micro-detail without halos.
Selective smoothing for flat regions while preserving structure on faces and text.
No upload, no servers. Your image stays on your device.
Push small references to wall-print resolution in seconds.
Photo, illustration, and portrait modes tune the pipeline per content.
This demo uses high-quality classical upscaling (bicubic + adaptive unsharp + selective denoise) and runs entirely in your browser. To swap in a neural model like Real-ESRGAN, you can drop in an ONNX runtime build — the UI is wired for it.
Browsers limit canvas size. We clamp the long edge so very large outputs don't crash the page.
Nowhere. Everything is processed on your machine — no upload, no telemetry.