Pixels, percentages, and when to use each
Exact pixel dimensions are the right choice whenever a destination has a hard requirement — a CMS that expects a 1200×630 social share image, an ad network banner spec, or a form asking for a specific photo size. Percentage resizing is better when you want to scale proportionally without doing arithmetic — “make this 50% smaller” is often what you actually mean when a photo just feels too large, without a specific target in mind. This tool supports both because they solve different problems: one is about hitting a spec, the other is about proportional scaling.
The math behind aspect-ratio locking
An image’s aspect ratio is simply width divided by height. When the lock is engaged and you change the width, the tool computes newHeight = originalHeight × (newWidth / originalWidth) and rounds to the nearest whole pixel, and the reverse for height changes. This guarantees the resized image is a uniform scale of the original — nothing stretches unevenly in one direction. Presets like 1920×1080 (16:9) are exact ratios; if your source photo isn’t already 16:9, forcing those exact dimensions without cropping will letterbox or distort the image, which is why presets in this tool set the target dimension pair directly rather than trying to force-fit unrelated source ratios — combine with the Crop tool first if you need an exact ratio from an image that doesn’t already have it.
Downscaling vs. upscaling: why one is safe and the other isn’t
Shrinking an image is a well-understood, essentially lossless-for-quality operation: the resizing algorithm (this tool uses high-quality bicubic/Lanczos-style resampling depending on the browser’s canvas implementation) averages groups of source pixels into each new, smaller pixel, which if anything reduces noise and produces a clean result. Enlarging an image is fundamentally different — the algorithm has to invent pixel values in the gaps between what was actually captured, using interpolation. This can look acceptable for modest enlargements (110-130%) but produces visibly soft or blurry results beyond that, because interpolation can only guess at detail, not create it. If you need a genuinely sharper large version of a small image, no resizing tool — this one included — can substitute for a higher-resolution source.
Common target sizes and why they exist
1920×1080 is the standard “Full HD” dimension used by most monitors, presentation templates, and video thumbnails — a safe default when you need “a normal-sized photo” without a more specific requirement. 800×600 is a legacy 4:3 standard still requested by some older forms and embedded systems. Social platforms each enforce their own preferred ratios (Instagram’s 1:1, 4:5, and 1.91:1; general Open Graph link previews at 1200×630) specifically because content that matches their expected ratio avoids an unpredictable auto-crop applied by the platform itself. Passport and ID photo sizes are set by government specification and combine a physical size with a required print resolution, which is why they’re expressed in exact pixels here rather than as a ratio.
A note on batch resizing
When you drop multiple images into batch mode, the same target dimensions or percentage apply to every file, but each file’s own aspect ratio (if locked) is respected independently — resizing five differently-shaped photos to “50%” gives each one its own correctly-scaled result rather than forcing them all into one uniform shape. This is the fastest way to prepare a whole folder of photos for a gallery, upload batch, or product listing without processing each one by hand.