Image Resizer
Resize any image with contain / cover / stretch fit modes.
Overview
The Image Resizer scales an uploaded picture to a new width and height with a choice of three fit modes — contain, cover and stretch. Pick your target dimensions and the tool produces a re-encoded copy that fills the target while honouring the chosen aspect-ratio rule.
This is the default tool for prepping a hero banner at a fixed width, generating responsive image breakpoints, producing OG card thumbnails, or shrinking an oversized phone photo before attaching it to an email. It also handles the perennial task of resizing an image online without losing quality.
How it works
Contain scales the image so it fits entirely inside the target box without cropping; the unused area is left transparent (or filled with a background colour). Cover scales so the target box is fully filled and crops whatever extends beyond it — the canonical "no letterbox" mode for thumbnails. Stretch ignores aspect ratio and forces the source into the target dimensions.
The resampling kernel is Lanczos with a 3-lobe support, which gives sharp results for both upscales and downscales while suppressing the moiré and aliasing of a naive nearest-neighbour. For very large downscales the pipeline first runs a box filter to roughly half-size before the final Lanczos pass — a standard trick that keeps the kernel cost reasonable while staying perceptually equivalent.
Examples
Before: 4000x6000 photo, target 1200x1200, fit=cover
After: 1200x1200 thumbnail, top and bottom cropped.
Before: 1920x1080 landscape, target 800x800, fit=contain, bg=white
After: 800x800 image with white bars on top and bottom.
Before: 600x900 portrait, target 1200x1200, fit=stretch
After: 1200x1200 image — squashed into a square aspect ratio.
FAQ
Will upscaling look soft?
Lanczos upscaling is sharp but cannot invent detail that wasn't in the source. For more than a 2x enlargement consider an AI upscaler — pure resampling will look soft at any large factor.
Which fit mode for OG / social cards?
Cover to a 1200x630 target produces a clean Twitter or LinkedIn card. Contain leaves letterboxing that most platforms then crop on top.
Does it preserve transparency?
PNG and WebP transparency is preserved through contain. Cover and stretch keep alpha unchanged; contain fills the unused area with the chosen background (transparent by default).
Will EXIF orientation be respected?
Yes — the tool applies the EXIF orientation first, then resamples. The output's orientation tag is reset to 1 so viewers don't double-rotate.
Is there a single-axis option?
Set the other dimension to 0 to scale by aspect ratio. Most users supply both dimensions explicitly to be precise.