Image Resize

Resampling and what it can and cannot do

Resizing an image means resampling it: computing a new grid of pixels from the old one. Reducing size is well-conditioned — several source pixels contribute to each output pixel, and the result is a faithful, slightly softer version of the original. Enlarging is fundamentally different. There is no additional information available, so the algorithm interpolates between known pixels, and the output is larger without being more detailed. This is why the upscale guard exists: the operation is not wrong exactly, but it is almost never what someone wants when they have not thought about it, and a silent enlargement leaves the user with a file that is bigger and worse.

Aspect ratio, and why distortion is easy to cause by accident

Aspect ratio is the relationship between width and height. Changing it stretches the content: circles become ellipses, faces become wide or narrow. It is easy to cause unintentionally by typing both dimensions from a specification without checking whether the source shares that shape. Fit within mode makes distortion impossible by construction, deriving whichever dimension you leave blank and, when you supply both, scaling the image to sit inside the box rather than filling it. Exact size mode does exactly what it is told, which is occasionally what you need and frequently not.

Batch behaviour and partial failure

A batch is processed one image at a time, and the queue reports progress as it goes. If a file fails part-way through — a corrupt image, an unsupported variant — everything that already succeeded remains available for download rather than being discarded along with the failure. The error band names the file that failed and why. This matters for large batches, where discovering a problem should not cost you the twenty files that worked.