Image Size Reducer
PNG upscaler

Upscale a PNG

Increase the pixel dimensions of an image by a set factor or to an exact size. Choose how the missing pixels are estimated, and the result is a bigger file whose edges and transparency survive the enlargement.

  • Scale by factor or exact size
  • Nearest, bilinear, bicubic, Lanczos
  • Alpha edges handled as alpha
  • Whole multiples for pixel art
  • Runs in the browser
Open the upscaler

What Enlarging Actually Does

Every new pixel in a bigger image is a guess. The useful question is which guess, and whether it suits the picture in front of you.

Upscaling is the one image operation that is asked to do something impossible, and it is worth being clear about that before starting. A smaller image contains a fixed number of pixels. A larger one contains more, and every pixel that was not in the original has to be estimated from its neighbors. No method invents detail that was never captured; the only choice is how convincingly the estimate is made. That does not make it useless. There are legitimate reasons to upscale a PNG, from print resolution to high-density screen assets, and with the right method the result holds up well. What it cannot do is turn a small thumbnail into a large photograph, and a page promising otherwise is selling a sharpening filter under another name.

What the Upscaler Does

Upscale a PNG by a whole multiple or to an exact size, with four resampling methods and alpha treated as alpha rather than as color.

Four resampling methods

Nearest for pixel art, bilinear and bicubic for photographs, and Lanczos when detail retention matters most.

Whole-multiple presets

Multiples of two, three, and four, which keep a pixel grid exact when the source is hard-edged artwork.

Exact target size

Enter a width and height when the destination has a fixed requirement rather than a multiplier.

Alpha-aware scaling

Color and transparency are resampled together, so cut-out edges stay clean instead of picking up a halo.

Local processing

The scaling runs in the browser and the image never leaves the device.

ImageSizeReducer

ImageSizeReducer

@imagesizereducer

Bigger is not sharper

A photograph that looks crisp at six hundred pixels does not become twice as detailed at twelve hundred. It becomes larger and slightly softer, because the interpolation averages what it has rather than revealing what it does not. Knowing this changes what you ask for. If the goal is a screen image that fills more space, the enlargement is fine and the softness is invisible at normal viewing distance. If the goal is to read small text off a screenshot, or to recover a compressed thumbnail, no amount of upscaling a PNG will help, and the time is better spent finding a larger original.
Upscale it
A small image shown enlarged and slightly softened
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

Choosing the method that matches the picture

The resampling method decides how the new pixels are estimated, and the right choice depends entirely on what kind of image it is. Nearest-neighbor simply copies the closest existing pixel into each new slot, which is exactly what pixel art and hard-edged graphics want, and visibly blocky on anything else. Bilinear and bicubic average across the neighbors and produce a smoother result, with bicubic the usual default for photographs. Lanczos goes further and holds detail and contrast better at the cost of a slight ringing near sharp edges. For most work a bicubic enlargement from the largest source available is the safe answer; for a sprite or an icon, nearest is the only answer.
Upscale it
Four resampling methods compared on the same detail
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

Whole multiples and pixel art

Multiplying by two, three, or four is a special case worth understanding. At an exact multiple with nearest-neighbor scaling, every original pixel becomes a neat block of identical pixels and nothing is estimated at all. A sprite scaled three times is a perfect, larger version of itself. Scale by one and a half instead and the blocks come out uneven, because a pixel and a half cannot be divided into whole pixels. Some columns end up two source pixels wide and some only one, which reads as a wobble. When you upscale a PNG for a sprite sheet or an interface asset, staying on whole multiples keeps the grid intact.
Upscale it
Pixel art scaled by an exact multiple and by a fractional one
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

What happens to the alpha channel

Transparency is where an enlargement can go visibly wrong, and it is specific to formats that carry an alpha channel. A fully transparent pixel still has color values stored beside it, and those values are arbitrary. Interpolating them as if they mattered drags stray color into the edge of the visible artwork. The symptom is a dark or colored halo tracing the outline of a cut-out, appearing only against certain backgrounds and only after scaling. Proper alpha-aware resampling handles color and transparency together rather than treating them as independent. It is a detail, and it is the difference between a clean cut-out and one that looks faintly dirty on a colored page.
Upscale it
A scaled cut-out showing a colored halo along its edge
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

What enlarging is genuinely for

There are real jobs for it. A print layout that needs roughly three hundred pixels per inch of printed size often requires an image larger than the screen version. A website that serves higher-density assets needs two or three times the display dimensions. A video or a presentation has a minimum size that a small asset does not meet. In each case the source is decent and the enlargement is modest, and that is the zone where you can upscale a PNG and get a good result. The mistake is treating the same operation as a rescue tool for an image that was already too small for the job it was given.
Upscale it
A modest enlargement used to meet a print requirement
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

A larger file is not a better one

A larger file looks like a better asset at a glance, and file size is a poor proxy for quality. Enlarging triples the pixel count while the actual information stays where it was, so the image is heavier to load and no more detailed. That trade is worth weighing before shipping anything. On the web, a doubled file that renders identically at display size is a cost with no benefit, and the honest options are usually to export at the needed size in the first place, or to accept the smaller source and let the layout work around it.
Upscale it
The same picture shown at two file sizes with equal detail
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

Repeated scaling compounds the softness

PNG is lossless, so unlike JPEG an enlargement does not add ringing or block artifacts to the file. What it does add is softness, and softness accumulates. Enlarging by two, then by two again, is not the same as enlarging once by four; each pass estimates new pixels from pixels that were themselves estimated, and the result is noticeably blurrier for the same final dimensions. So the rule is to scale once, from the largest source available, at the final size actually needed. If you have to upscale a PNG in stages because a tool imposes a limit, go as far as possible in a single step and treat any further passes as a last resort.
Upscale it
An image enlarged in one step beside one enlarged twice
ImageSizeReducer

ImageSizeReducer

@imagesizereducer

Trying a factor before committing

Enlargement is often a quick check as much as a final step: does this asset hold up at two times, or does it need to be re-exported at source? The scaling runs on the copy already on the machine. Nothing is uploaded, no account is needed, and both the dimensions and the method can be changed as often as the question takes to answer.
Upscale it
An image being scaled inside a browser window

How to Upscale a PNG

Four steps. Step three is the one that decides whether the enlargement looks clean or blocky.

Step 1

Upload the largest source

Start from the biggest original you have, because a later enlargement will not recover what an earlier one lost.

Step 2

Choose a factor or a size

Whole multiples suit pixel art and hard-edged graphics; an exact width and height suits a fixed layout slot.

Step 3

Pick the resampling method

Nearest for flat artwork, bicubic for photographs, and Lanczos where fine detail has to hold up.

Step 4

Inspect at full zoom

Check the result at one hundred percent, and on both a light and a dark surface if the image has transparency.

Who Enlarges PNG Files

Four kinds of work where a file has to be larger than the source that was handed over.

PD

Print designers

prepress sizing

A supplied asset is often a little short of the resolution a layout needs. A modest enlargement from a decent source bridges the gap without a reshoot.

FD

Front-end developers

high-density assets

Serving a two-times version means every icon and illustration needs double the pixels. Whole multiples keep the sharp ones sharp.

GM

Game artists

sprites and tiles

Pixel art only survives enlargement with nearest-neighbor scaling at exact multiples. Anything else smears the grid the whole style depends on.

RT

Retouchers

client files

The usual request is to make a file bigger, and the usual answer is that bigger is not better. Setting that expectation early saves a lot of wasted effort.

Frequently Asked Questions

How much detail can be recovered, and why the method matters more than the multiplier.

Make it larger

Upload a PNG, pick a factor or an exact size, and choose how the new pixels are estimated, all without the image leaving the browser.

Upscale it
More Pixels, Same Information

Enlarging is estimation, and the method is the whole decision

Upscaling asks an image to become something it is not. A small file holds a fixed amount of information, and a larger one needs more pixels than that information can fill, so every new pixel is an estimate drawn from the ones around it. No method breaks that rule; the only variable is how convincing the estimate looks. That is why the resampling choice matters more than the multiplier. Nearest-neighbor copies the closest pixel and is exactly right for sprites, icons, and any graphic built from flat blocks, while being visibly crude on a photograph. Bicubic and bilinear average across neighbors and suit photographic content, and Lanczos holds fine detail hardest at the cost of a faint ringing along sharp edges. Transparency adds a second decision that has nothing to do with the first. Fully transparent pixels still carry color values, and interpolating those as though they meant something drags stray color into the edge of a cut-out, which is where the halo around a scaled logo or icon comes from. Handle alpha together with color and the problem does not arise. Finally, scale once. Each pass estimates from pixels that were already estimated, so enlarging twice by two is blurrier than enlarging once by four at the same final size. Do all of that and, when you upscale a PNG, the result is a larger file that is honestly larger rather than pretending to be more detailed than its source ever was.

Smaller files, same picture.