"Compress without losing quality" is the phrase everyone searches for and no tool can deliver, because compression is lossy by definition. What you can do is decide in advance how much quality you are willing to trade, and then stop before the trade becomes visible. That is a decision, not a setting, and getting it right takes about four minutes.
Here is the order that actually matters. Most people start at step three and never reach the step that makes the largest difference.
What compression actually takes away
A JPEG does not store your picture. It stores a mathematical description of your picture, and it throws away the parts a human eye is least likely to miss: fine detail in busy areas, subtle differences between neighbouring pixels, and colour precision that the eye resolves poorly anyway.
At high quality settings it discards almost nothing and you cannot tell. As the setting drops it starts discarding detail you can see — first in textured areas like grass and hair, then as blocky patches around hard edges, then as flat colour banding in skies.
So the honest goal is not "no loss". It is "loss that stays below the threshold where anyone notices, at the size you will actually view the image".
Step one: decide where the image is going
This determines everything else, and skipping it is why people either over-compress or under-compress. The same photograph has four completely different targets.
- A website or blog post. Roughly 1200 to 1600 pixels on the long edge. Quality 75 to 82. Target under 250 KB.
- A marketplace or shop listing. Usually a hard minimum on the long edge — often 1600 or 2000 pixels — and a hard maximum on file size. Check the requirements before you start.
- Email or a messaging app. Nobody views these full-screen on a large display. 1000 pixels and quality 70 is generous.
- Print. Different rules entirely. You want pixels per inch of paper, not file size, and compression is the wrong lever. Nothing below 2000 pixels will look right at A4.
Write your target down. A number you decided before you started is what stops the fiddling.
Step two: choose the format before the quality slider
Format choice affects file size far more than the quality slider does, and it is usually decided by accident.
- JPEG. The default for photographs. No transparency. Universally supported, including by software from 2004.
- PNG. Lossless, supports transparency, and produces enormous files for photographs. Correct for screenshots, logos and anything with flat colour or text.
- WebP. Roughly 25 to 35 percent smaller than JPEG at equivalent visual quality, with transparency support. Supported by every current browser.
If you are publishing to a modern site, WebP is usually the right answer and JPEG is the safety net. If a photograph is being turned into a PNG, the size will surprise you and the cause is the format, not the quality setting.
Step three: resize first, then compress
This is the step that does the heavy lifting, and it comes before the quality slider rather than after it.
Halving the pixel dimensions of an image reduces the number of pixels to four times fewer — 4000 by 3000 becomes 2000 by 1500, and the file drops to roughly a quarter of its size before any quality loss at all. Compressing the same image from quality 90 to quality 60 reduces it by perhaps forty percent, with visible damage.
So resize to the dimensions you actually need, and only then reach for the quality slider. An image that is 4000 pixels wide when the page renders it at 800 is carrying four times the data anyone will ever see, and no amount of compression fixes a mistake of that shape.
Step four: set quality by looking, not by number
There is a usable range and it is narrower than the slider suggests.
- Below 60. Blocking around edges becomes obvious on any display. Avoid unless the image is a thumbnail.
- 70 to 85. The practical zone for the web. Most photographs are indistinguishable from the original in this band at normal viewing size.
- Above 90. File size climbs steeply for a difference nobody can see. Rarely worth it outside print preparation.
The reliable method is to sit at 80, look at the result at 100 percent zoom on the sharpest area of the image, and move down in steps of five until you can just start to see a change. Then go back up one step. That is your number for this image, and it will differ from the last one — a photograph of foliage tolerates far less compression than a flat studio backdrop.
Step five: check the edges and the sky
Two places fail first, and both are easy to miss on a small preview.
Hard edges against a plain background — a product on a white sweep, a dark silhouette against a bright wall — are where blocking appears earliest, because the compressor has to spend data describing a sudden transition.
Gradients — a blue sky, a soft shadow on a wall — are where banding appears, as smooth transitions break into visible steps.
Zoom to 100 percent, look at both, and only then accept the result. A thumbnail preview hides exactly the artefacts you are trying to avoid.
What not to do
Four habits cause most of the unnecessary quality loss people blame on tools.
- Re-compressing the same file repeatedly. Every save from a JPEG introduces a fresh generation of loss. Edit from the original, always, and save once.
- Screenshotting a photograph. This converts it to screen resolution and re-compresses it. The result is always worse than compressing the original.
- Upscaling a small image to hit a size requirement. Upscaling invents pixels; it does not recover them. Starting from a larger original is the only real fix.
- Compressing before cropping. Crop first so you are compressing only the part you keep, instead of spending data on the part you are about to throw away.
The image compressor runs entirely in your browser, which is worth knowing if you are working with client material: resize, set quality, check at 100 percent, and the file never leaves the machine. If your source is a screenshot or a logo rather than a photograph, converting it to WebP will usually beat compressing a PNG — the format was the problem, not the quality setting.