Compress Image to Exact Target Size — Free Online Tool

How binary-search compression works

No image encoder accepts "make this 100 KB". JPEG and WebP take a quality value between 0 and 1, and the number of bytes that falls out the other end depends entirely on the picture — a flat blue sky at quality 0.6 might be 40 KB while a crowd scene at the same setting is 400 KB. The only way to find the right setting is to try one and measure.

Trying them one at a time is slow. Bisection is not. The tool encodes at the midpoint of the remaining range, checks the result against your target, and throws away the half of the range that cannot possibly contain the answer:

Each attempt halves the search space, so eight encodes pin the best quality to within half a percent. A linear walk would need well over a hundred to do the same, and every one of them costs a full encode of your image.

There is a floor to this. If your target is so small that even quality 0.02 overshoots — 10 KB from a 24-megapixel photo, say — no quality setting can save it, because you are paying for pixels rather than for detail. At that point the tool starts reducing the dimensions instead, estimating the first step from the measurement it already has (bytes track pixel count closely), then re-running the quality search at the smaller size so the picture claws back everything the shrink paid for.

Compression vs. resizing: what is the difference?

They are two different levers and they fail in different ways. Resizing changes the pixel dimensions — 4000×3000 becomes 1200×900 — and you lose detail you could only have seen by zooming in. Compressing keeps every pixel but stores them less precisely, and you lose detail you would only notice by looking closely: blocking around hard edges, mottling in smooth gradients.

For a fixed byte budget there is a trade between them, and which one to spend first depends on what the image is for. A passport photo that will be printed at 35×45 mm should keep its quality and lose pixels. A web hero image that will be displayed at 1200 px wide should lose the pixels it will never show, then keep as much quality as the budget allows. That is what the Priority control does — it decides which lever the search pulls first.

Why your 100 KB might not be their 100 KB

"KB" means two different things depending on who wrote the upload form. Some validators check size <= 100000 (decimal), others size <= 102400 (binary, which is what Windows shows you in Explorer). A file that satisfies the second can fail the first, and the form will usually just say "file too large" without explaining which one it meant.

This tool reads your target the strict way — 1 KB = 1 000 bytes — and aims a little under that. The result clears both kinds of validator, at the cost of a fraction of a percent of quality you were never going to see.

Which output format should you pick?

Five are offered, and the choice matters more as the target gets tighter — at 500 KB they all look the same, at 20 KB they do not.

For the three lossy formats the catch is acceptance, not support. Every current browser reads WebP and AVIF, but plenty of government portals, older recruitment systems and print shops still reject the extensions outright. The rule of thumb: WebP or AVIF for anything that ends up on a website you control, JPG for anything you are handing to someone else's form. JPG is the default here for exactly that reason. AVIF is also much slower to encode — the first image downloads a one-time encoder — so it suits a handful of images rather than a batch of fifty.

PNG behaves differently from the rest. It is lossless, so there is no quality dial to turn and the only way to make it smaller is to remove pixels. The tool therefore searches the dimensions instead, and the Priority control disappears because there is no quality left to prioritise. Expect a small target to shrink a PNG hard — 20 KB of lossless photo is roughly a 90-pixel-wide picture. If you want to keep the original dimensions of a PNG and simply make the file smaller, the PNG compressor is the better page. Choosing a lossy format for an image with a transparent background is flagged before anything is flattened onto white.

Why browser-based compression is more private than cloud tools

Most "compress to 100 KB" sites work by uploading your file, compressing it on a server and sending it back. That means a copy of your passport scan, your salary slip or your child's school photo exists on somebody else's disk, covered by whatever their retention policy happens to say this month.

Resizo has no backend that receives images. The search runs in your tab using the browser's own canvas encoder, which is the same code path your phone uses to save a photo. Two ways to confirm it: open the Network panel in your browser's developer tools and watch that nothing is sent while you compress, or simply disconnect from the internet after the page has loaded and compress anyway — it still works.

A useful side effect: re-encoding drops EXIF metadata, so GPS coordinates, camera serial numbers and timestamps do not travel with the compressed copy. That is usually what you want when a file is going to a stranger. If you need the metadata, keep the original.