Resize image to target file size

Enter the source width and height, choose JPEG, PNG, WebP, or AVIF, and provide the maximum size in decimal kilobytes. The calculator returns suggested dimensions, quality, estimated size, and the smallest realistic estimate. It works from metadata rather than image pixels, so its result is a planning estimate, not a promise about a particular encoder or photograph. The same inputs always produce the same answer.

Turn a file-size limit into practical export settings

Start with the image's actual pixel width and height, its intended output format, and the maximum file size accepted by the destination. The calculator first estimates a normal export at a format-specific reference quality. If that estimate already fits, it preserves the original dimensions and recommends no change. For lossy JPEG, WebP, and AVIF output, it next lowers the modeled quality while keeping every pixel. When quality alone is unlikely to be enough, it holds a conservative minimum quality and calculates a proportional scale for both dimensions. PNG is treated as lossless, so it recommends scaling rather than pretending that a JPEG-style quality slider controls its size. The output includes the chosen action, integer output dimensions, scale percentage, quality, estimated file size, original estimate, and minimum estimate. Use these values as the first settings in an image encoder, then measure the encoded result and make a small final adjustment if the receiving service enforces a strict byte limit.

Understand what the estimate can and cannot know

File size cannot be derived exactly from dimensions and format alone. A flat icon, a noisy night photograph, and a detailed illustration can share the same width, height, and format yet compress to very different sizes. Encoder implementations, chroma subsampling, metadata, animation, alpha transparency, color profiles, and optimization settings add more variation. This calculator therefore uses a transparent analytic model: each supported format has a typical number of compressed bytes per pixel, a small container overhead, a reference quality, and, for lossy formats, a smooth quality curve. The returned confidence is deliberately labeled as a rough estimate. It does not inspect, upload, decode, or alter the image, and it never claims to produce an exact file. For reliable production automation, encode once with the recommendation, compare the resulting bytes with the target, and use an encoder's bounded search to converge. The estimate is most valuable before encoding, when you need a sensible initial quality and resolution instead of guessing blindly or repeatedly trying unrelated settings.

Handle very small targets without unusable recommendations

An arbitrarily tiny target is not realistic for every image. Even an empty image file needs headers and container data, and shrinking a large photograph to a nearly invisible bitmap is rarely a useful interpretation of resize. The calculator establishes a practical floor by preserving the source aspect ratio while allowing the shorter side to fall no lower than 32 pixels, then applying the format's minimum modeled quality. If the requested target is below that estimate, the request fails with an invalid-input error that states the calculated minimum in kilobytes. This is preferable to returning zero-sized dimensions, negative arithmetic, or an apparent guarantee that no ordinary encoder could meet. Decimal kilobytes are used, meaning one KB is modeled as 1,000 bytes. Dimensions must be positive integers, the total source area is bounded, and output dimensions are rounded down to whole pixels so the estimate stays at or below the modeled target. Always preserve a safety margin when an upload portal rejects files at an exact boundary.