Resize any image to exactly 1024x1024 pixels. The default square output size for Midjourney, DALL-E, Stable Diffusion and many AI tools, plus app icons and profile artwork.
Last updated: July 2026
1024x1024
Exact Size
Fill, Fit, Stretch
Modes
JPEG, PNG
Output
Private
No Uploads
Dimensions fixed to 1024 x 1024 px for this page.
Scales and crops to fill exact size. No white space. No distortion. Recommended.
Drop image here to resize to 1024x1024
JPEG, PNG, WebP, AVIF, HEIC. All processing stays in your browser.
If your image is smaller than 1024x1024 pixels, the tool will upscale it and show a warning above the download button. For the sharpest result, start with a source image that is at least 1024 pixels on its shortest side.
Drop or select your image
JPEG, PNG, WebP, AVIF or HEIC. Dimensions are already locked to 1024x1024.
Choose Fill, Fit or Stretch
Fill crops to cover, Fit pads with white, Stretch forces the exact size.
Choose JPEG or PNG
PNG keeps transparency, JPEG gives a smaller file.
Check the live preview
Change settings and the result updates without re-uploading.
Download your 1024x1024 image
Nothing is uploaded, the file saves straight to your device.
Common reasons people resize to 1024x1024
1024x1024 is the de facto standard square resolution for AI-generated images and is also widely used for icons and profile artwork.
AI image generators
Midjourney, DALL-E and Stable Diffusion commonly default to 1024x1024 for square output. Resize reference images to this size before uploading them as prompts.
App icons and product artwork
1024x1024 is the master size many app stores request before generating smaller icon sizes automatically.
NFT and digital art
A common export size for digital artwork and collectible images shared on marketplaces and social platforms.
Source file for smaller icons
Because 1024 divides cleanly into 512, 256 and 128, it works well as a master file for generating a full icon set.
1024x1024 is one of those numbers that shows up across completely different fields for the same underlying reason, it's a power of two. Specifically it's 2 to the power of 10, which means it divides evenly all the way down through 512, 256, 128, 64 and 32. Graphics processors, texture pipelines and neural network architectures all prefer power-of-two dimensions because they map cleanly onto how memory is allocated and how convolution operations are structured internally.
The practical consequence is that when you feed a reference image into an AI image generator, it typically gets rescaled to the model's native resolution before processing begins. If your reference image is 3000 by 4000 pixels from a phone camera, it's going to get squashed to roughly 1024 on the shortest side anyway. You get more control over how that squashing happens if you do it yourself first, with a mode you've chosen, rather than letting the model's preprocessing handle it.
Resizing a non-square image to a square canvas means making a choice about the mismatched pixels. Fill scales the image to completely cover the 1024x1024 canvas and crops the center as needed, no padding, no distortion, and it's the default mode here. Fit scales the image to fit entirely inside the canvas with nothing cropped, filling any remaining space with white, which means it's not the right choice if you need a transparent background. Stretch distorts the image to exactly match 1024x1024, which almost always looks wrong for photos but can work for logos or abstract textures.
For AI reference images, Fill is usually the better starting point, diffusion models tend to respond better to a frame that's filled rather than one with blank padding around the edges. If you need precise control over what gets cropped rather than relying on Fill's automatic center crop, the crop tool lets you set an exact 1:1 crop and position it manually before you resize.
Because 1024 is exactly double 512, and 512 is double 256, and 256 is double 128, you can downscale a 1024x1024 master to any of these sizes without fractional pixel rounding. This is why a 1024x1024 source is a common starting point for app icon submissions, smaller sizes get generated from that master. If you're building a favicon and want to go from 1024x1024 down to a proper multi-size favicon.ico, converting the resized PNG with the PNG to ICO converter is the next step, or use the favicon generator to produce a full set of sizes at once. If you need a smaller file for the web afterward, converting the PNG to WebP is worth doing, since this page's own output is limited to JPEG or PNG.
Using Fit mode for an AI reference image
Fit adds white padding around your image to avoid cropping. Diffusion models generally respond better to a frame that's fully filled, so Fill mode is usually the better default for AI reference uploads.
Expecting Fit mode to give a transparent background
Fit always fills the padding with solid white, even when you've chosen PNG output. If you need real transparency around your subject, use Fill or Stretch with a source image that already has a transparent background.
Upscaling a small image by a large amount
Going from something like 300x300 to 1024x1024 asks the tool to invent detail that was never in the original. The result will look soft. Start from the highest resolution version you have whenever possible.
Choosing JPEG when you need a transparent app icon
JPEG doesn't support transparency at all. If the icon or artwork needs a transparent background, output as PNG instead.
Assuming you'll get a WebP or AVIF file
Output on this page is JPEG or PNG only. If you specifically need a WebP or AVIF version, convert the downloaded file afterward with a dedicated converter.
Not checking the upscale warning
If your source is smaller than 1024x1024 in either dimension, the tool upscales it and flags this above the download button. It's easy to miss if you're moving fast, worth a second look before you use the result somewhere quality matters.
1024x1024 pixels is roughly one megapixel, close to the native output resolution most diffusion based AI image generators are trained on. Midjourney, DALL-E and Stable Diffusion commonly default to 1024x1024 for square outputs. Generating at much higher resolutions directly can introduce artifacts, so many workflows generate at 1024x1024 first and upscale afterward if a larger size is needed.
The tool upscales it to 1024x1024 and shows a warning above the download button once the result is ready. Upscaling stretches existing pixels rather than adding new detail, so the output may look softer than a native 1024x1024 image. For the sharpest result, start with a source image that is already at least 1024 pixels on its shortest side.
Exactly 1024x1024. The width and height are locked to this page's preset, so whichever mode you choose (Fill, Fit or Stretch), the output canvas is always precisely 1024 by 1024 pixels.
Fill is the default and scales your image to completely cover the 1024x1024 square, cropping the center as needed, with no distortion and no padding. Fit scales the image to fit entirely inside the square without cropping, adding white padding on the sides that don't match. Stretch forces the exact 1024x1024 size by distorting the image if the aspect ratio doesn't already match, and almost always looks wrong for photos. For AI reference images, Fill is usually the better starting point since it avoids blank padding.
Yes, if you choose PNG output and use Fill or Stretch mode, transparency from your source image carries through. Fit mode is the one exception, it always fills the padding around your image with a solid white background, even when the output format is PNG, so avoid Fit if you need a fully transparent 1024x1024 canvas.
You can upload JPEG, PNG, WebP, AVIF or HEIC. The output is JPEG or PNG only, there's no WebP or AVIF export on this page. If you need a smaller WebP or AVIF file afterward, convert the downloaded PNG or JPEG separately.
No. The image is decoded to raw pixels and re-encoded during resizing, so EXIF metadata, including GPS location, is not carried over to the 1024x1024 output file.
No. Resizing runs entirely in your browser using WebAssembly. Your image is never sent anywhere.
This page resizes one image at a time. If you're preparing a batch, for example a full icon set, you'll need to run each image through separately here.
The quality slider defaults to 85, which is a good balance for most AI reference images and icon exports. Drag it lower for a smaller file, or higher if you're feeding the result into another tool that's sensitive to compression artifacts. PNG output has no quality slider, since it's always encoded losslessly.
Shrinking a larger image down to 1024x1024 rarely causes visible quality loss. Enlarging a smaller image up to 1024x1024 can soften it, since the tool has to interpolate pixels that were never in the original.
Fill mode automatically crops from the center, which works well most of the time. If you need to choose exactly what stays in frame, for example keeping a face or logo off-center, use the crop tool first to set a 1:1 crop, then resize the cropped result here.
1024x1024 is also a common size for high resolution app icons, NFT artwork, large profile pictures and master design files that get exported down to smaller sizes. Since it's exactly double 512x512, it downscales cleanly to 512, 256 or 128 pixels without fractional rounding.
No sign-up. No uploads. Drop your image above and download the 1024x1024 result in seconds.
