Upscale

Upscale enlarges an image in your browser by doubling it repeatedly rather than stretching it in one draw, which keeps edges smooth instead of blocky. A browser asked to quadruple an image in a single call has too few source pixels under each destination pixel; two extra draws give its resampler a full neighbourhood every time. Enlarging cannot invent detail that was never captured — an optional 3×3 unsharp mask runs once after the last step to put some edge contrast back, and that is the whole of it.

How it works

  1. Drop your images Drag one or more images in. Each is decoded and its enlarged size is worked out for you.
  2. Choose a scale 2×, 3× or 4×, or Custom for exact pixel dimensions. Clear one side of the custom pair to keep the aspect ratio of the source.
  3. Read the step ladder The intermediate sizes the image will be drawn through are listed before you run anything. No step is more than double the one before it, and the last entry is exactly the size you asked for.
  4. Set sharpening A 3×3 unsharp mask runs once after the final enlargement step, at 35% by default. Above 80% it starts drawing visible halos along hard edges, where a dark shape meets a light one.
  5. Apply and download Pick keep, PNG, WebP or JPEG and apply. One file downloads on its own; a batch comes back as a ZIP.

Questions

Is this AI upscaling?
No. There is no model and no reconstruction — the tool resamples, drawing the image through a ladder of doubling steps so the browser's resampler always has a full neighbourhood of source pixels under each output pixel. Enlarging cannot recover detail that was never captured; better resampling makes the result smooth rather than blocky, and nothing more.
Why double repeatedly instead of scaling in one go?
A single draw to 4× leaves each destination pixel with too few source pixels beneath it, and edges come out blocky. Doubling gives the resampler a full neighbourhood on every pass, for the cost of two extra draws. It is the mirror of the halving steps file4edit already uses when shrinking an image.
Is there a size limit?
There is no hard cap, but a target past about 80 megapixels is worth thinking about: the browser holds the enlarged image as raw pixels while it works, so a large batch can run out of memory. The tool shows the megapixel count and a RAM estimate before you run it, and warns you if the target is no larger than the source.

Every tool