Upscale images with AI.
Enlarge small or pixelated photos with a fast canvas resampler or an AI super-resolution model that rebuilds detail. Both run on your own device. Your image is never uploaded, and the result has no watermark.
Upscale an imageUpscale
How it works.
Three steps, no setup, nothing uploaded.
Drop a small or blurry image
Old photos, low-resolution downloads, screenshots, logos. Anything that needs to be bigger than it is.
Choose Fast or AI, and a factor
Fast resamples in multiple passes with a sharpening pass, instantly. AI runs Swin2SR, a super-resolution model that reconstructs detail; the first use downloads the model once.
Download the enlarged result
Up to 4× the original size. The result downloads as a PNG with no watermark.
Questions.
Does the AI upscaler run on a server?
No. The Swin2SR model runs in your browser. The first AI upscale downloads a ~21 MB model once, then it is cached and everything happens on your device.
When should I use AI instead of Fast mode?
Use AI for pixelated, low-resolution, or JPEG-degraded photos. It rebuilds detail that is not there. Fast mode is instant and fine for a clean image that just needs more pixels.
Will the AI invent details that were never there?
A little, yes. That is how super-resolution works: it generates plausible texture (skin, fabric, foliage) consistent with the image. For photos this looks natural; for tiny text or distant faces, treat the output as a reconstruction, not evidence.
How much can I enlarge an image?
Up to 4×. AI mode caps the input at 512 px on the long side to stay inside browser memory, producing up to a 2048 px result. For larger sources, Fast mode handles the full resolution.
Why did my first AI upscale take so long?
That was the one-time model download, not the upscale. Once cached, an AI pass typically takes seconds, and Fast mode is instant every time.