Honest limits
What This Resizer Will Not Do
To resize a logo without sending the file to a server, load the mark in the browser and change only its pixel dimensions. The workbench fits the source into the target frame with the aspect ratio preserved, centers it, and fills any remaining space with transparent or chosen-color padding. Contain mode keeps every part of the mark visible and is the right default for wordmarks, lockups, and any logo with a wide or tall silhouette. Cover mode crops the long edge to fill the frame, which is useful for a square avatar built from a horizontal logo but does hide the sides. Padding is a percentage of the shorter output side, so a 1080 pixel square at 6 percent gets a 65 pixel inset on each side. Corner radius is optional and suits round profile frames. Export is PNG, JPEG, or WebP. A generative image API cannot lock an exact pixel box, a transparent margin, or the promise that a mark is not redrawn, which is exactly why this page uses a local geometry compositor instead of a text-to-image model.
This page does not upscale a tiny raster logo into a crisp new mark, remove a background, vectorize the artwork, or invent a new logo concept. Background removal lives on Background Remover, and a fresh logo idea lives on AI Logo Generator. It also does not stretch a non-square logo to fill a square frame; forcing that match would distort the mark, so the frame is padded instead.
Canvas resizing is arithmetic, not a redraw. Downscaling a detailed logo to a 32 pixel favicon can still lose thin strokes, and upscaling a small source to a large frame can look soft because the pixels are interpolated. Files over 5 MB, sources outside 16 to 4,096 pixels per side, sources over 16 megapixels, and target sides outside 16 to 4,096 pixels are rejected before a preview is drawn. SVG upload is not supported; flatten the vector to a large PNG first so the resize starts from clean pixels.