Sunil Kapoor

My attention goes to the two decisions that are made once and then applied to a hundred files. Aspect ratio is the first. A resize that forces a source image into a fixed frame either crops or distorts, and both are visible. The choice of which depends on the destination: a profile photo wants a centre crop, a banner wants letterboxing or a regenerated design, and a product photo wants neither. Deciding once and applying it to everything is how a batch job produces two hundred files with the subject cut off. Upscaling is the second. A larger image cannot be recovered from a smaller one, so enlarging produces a file with more pixels and no more detail. It can still be the right choice for a fixed layout, and I cover when it is acceptable versus when it is a way of pretending a missing asset exists. Resampling method matters more at large reductions than people expect. Averaging and similar methods avoid the aliasing that nearest-neighbour produces, and aliasing shows up as jagged edges that look like a sharpening problem and are not. Sharpening should come after resizing, never before, because sharpening a small image amplifies the very artefacts resizing introduces. Colour profiles need deciding once too. Resampling in the wrong space changes the appearance of the result, so a profile-aware resize and a naive one produce visibly different images from the same source. I end every page on output naming, because a batch that overwrites its own source has no undo.