Compress an image to a target size

Pick a target and Image searches for the best quality that fits. It first tries high quality (95%); if that already fits, you get the best-looking result. Otherwise it encodes at the lowest allowed quality and binary-searches between the two, keeping the largest file that is still under the target — the highest quality that fits.

Keep dimensions only lowers quality (to as little as 5%). Allow resize if needed keeps quality at 60% or above and makes the image smaller in pixels instead, which looks far better than heavy compression. Image tells you exactly what it did.

Sizes are decimal: 1 KB = 1,000 bytes. That's the stricter reading of a limit, so a file made for “500 KB” fits whether the website counts 500,000 or 512,000 bytes. If the target can't be reached, you see the smallest result and why — it is never reported as a success.

Target 500 KB (500,000 bytes) for a 3.6 MB, 4000 × 3000 JPEG with Keep dimensions:

1. Quality 95% → 2.9 MB — too big.
2. Quality 5% → 180 KB — fits, so the answer lies between.
3. Quality 50% → 540 KB — too big. Quality 27% → 380 KB — fits.
4. Quality 39% → 470 KB — fits and is within 8% of the target, so the search stops.

Result: 470 KB at quality 39%, 4000 × 3000. With Allow resize the same photo would instead keep quality ≥ 60% at roughly 2600 × 1950 px. (Exact byte counts depend on the photo.)

• Already under the target: you're told, and get a metadata-free copy that still fits.
• Impossible without resizing: reported as not reached, with the smallest size achieved.
• Transparent images: WebP keeps transparency; JPEG asks for a background colour.
• PNG output can't be size-targeted losslessly, so targets produce JPEG or WebP.
• Tiny targets with resizing stop at a 16 px edge rather than shrinking forever.
• Cancel stops the search immediately.