Compress to Target Size

Compress to Target Size works backwards from a number. You give it a budget, and it searches the encoder's quality scale for the highest setting whose output still fits underneath: the first candidate is written at quality 80, and each following one halves the range that is still in play, between a floor of 5 and a ceiling of 95. That search runs to eight encodes, or stops sooner once the remaining range is under two quality points. Nothing here is estimated — every candidate is genuinely encoded and its bytes counted, so the size quoted in the result is the size of the file that lands in your downloads.

Quality is not always enough on its own. When the lowest quality tried is still over budget and “Shrink the dimensions” is left ticked, the picture is redrawn at 80% of its width and height and the whole quality search runs again. There are eight of those passes in all, so seven shrinks at most, which bottoms out at roughly a fifth of the original width — a photograph that still does not fit by then is handed back at that size rather than shrunk further. Smaller pictures stop before that, because no side is taken below 64 pixels. Three output formats are offered — JPG, WebP and AVIF — and PNG is deliberately missing from that list, because PNG is lossless, has no quality dial to turn, and therefore cannot be aimed at a size budget at all.

A budget can simply be impossible, and this is where the tool has to be believed rather than admired. If the passes run out at the lowest quality with the file still above your figure, what comes back is the smallest version it managed, the size that version actually reached, and a plain “no” against the target — not a quiet miss. Two other behaviours are worth knowing in advance. An image already under the budget in the format you asked for is handed back byte for byte, because re-encoding a small file can easily make it larger — and because those bytes are untouched, that copy keeps whatever EXIF it arrived with. Everything that is genuinely re-encoded is redrawn on a canvas first, so camera model, capture date and GPS position are dropped from those results.

The work happens inside this browser tab, through the same canvas encoder the browser uses to write images anywhere else, so the picture is not uploaded and the page keeps working with the network switched off. That is also where the boundary sits. The browser has to be able to decode what you feed it — HEIC photos from an iPhone cannot be decoded by any browser, and the HEIC Converter handles those on the server instead — and it has to be able to write the format you picked. AVIF writing only exists in recent browsers; where it is absent the job stops and asks you to choose JPG, rather than handing you a mislabelled file.

Runs on your device. Your file is processed directly in this browser.