What resizing an image actually does
An image is a grid of pixels with a fixed width and height. Resizing changes how many pixels are in that grid: downscaling throws away pixels using a resampling algorithm that blends neighbours so edges stay smooth, while upscaling adds new pixels by interpolating between the ones you already have. The visible content is the same picture — it just occupies more or fewer pixels, which is what controls both the file size and how the image displays on a given screen or print.
Most people resize for a concrete reason: an upload form rejects anything over a certain width, a web page loads slowly because the image is far larger than the space it fills, a social platform wants an exact canvas, or a print shop needs a specific pixel count at a specific density. This tool covers all of those with three ways to set the target — exact dimensions, a percentage of the original, or a named preset — plus a DPI control for print and a choice of resampling algorithm.
Dimensions, percentages, and presets
Dimensions mode is the most direct: type the width and height in pixels. Keep the aspect-ratio lock on and the two fields move together so the image is never distorted; turn it off when you need a precise canvas and don't mind a slight stretch. Percentage mode is the quickest way to halve or double an image — and because it scales each file by its own size, it is the right choice for a mixed batch where every image should shrink proportionally rather than to one shared size.
Presets remove the guesswork for the sizes that come up constantly. Instead of remembering that an Instagram story is 1080×1920 or a YouTube thumbnail is 1280×720, pick the platform from the list and the canvas is set for you. The catalogue spans the major social networks, common web resolutions like 1080p, 1440p, and 4K, and a few utility sizes such as a 512px favicon and an A4 page at 300 DPI for print.
Choosing a resampling algorithm
Resampling is the maths that decides each new pixel's colour, and the right algorithm depends on the image. Lanczos is the default because it gives the sharpest, most faithful result for photographs and detailed graphics; it uses a windowed sinc filter that preserves fine detail without the softness cheaper methods introduce. It runs in a background WebAssembly worker so a large image doesn't freeze the page while it processes.
Bicubic is a smooth, fast interpolation handled directly by your browser — an excellent general-purpose choice that is virtually indistinguishable from Lanczos for most everyday resizing. Nearest neighbour is the specialist: it copies the closest pixel with no blending at all, which would look blocky on a photo but is exactly what you want for pixel art, game sprites, or any image where crisp, un-smoothed edges are the whole point of scaling it up.
DPI and print workflows
DPI — dots per inch — is one of the most misunderstood settings in image editing. It is not a property of the pixels themselves; it is a number stored in the file that tells print software how densely to lay those pixels onto paper. A 3000×2400 image is the same on screen whether it is tagged 72 or 300 DPI, but a print shop reading 300 DPI will produce a crisp 10×8 inch print, while 72 DPI would spread the same pixels across a much larger, coarser area.
For anything destined for the web, 72 DPI is the convention and the value is largely ignored by browsers, which care only about pixel dimensions. For print, 150 DPI is fine for posters viewed at a distance and 300 DPI is the standard for photos and brochures. When you set a DPI here, we embed it into the exported file's metadata — a pHYs chunk for PNG, the JFIF density fields for JPG — so the print pipeline reads exactly the density you intended.
Everything runs in your browser
Most online resizers upload your images to a server, process them in the cloud, and send the result back. Your private photos and confidential documents pass through someone else's infrastructure, where they may be logged, cached, or retained. OpusImg works differently: resizing happens entirely on your device. Lanczos uses a WebAssembly build of the same high-quality resampler behind Squoosh, and bicubic and nearest-neighbour use your browser's own canvas.
When you drop a file, it is decoded, resized, re-encoded at high quality, and handed straight back to you without a single byte being uploaded. There is no server to trust, nothing to delete afterward, and the tool keeps working even if you go offline. Batch resizing runs the same local pipeline across every file and bundles the results into one ZIP — fast, private, and free, the way image resizing should be.