How much can you compress an image before it shows?

The honest answer: about 80 percent

On a typical photograph you can throw away roughly 80 percent of the file before anyone notices anything, and often more. A 4 MB camera JPEG saved at quality 80 usually lands somewhere between 400 KB and 900 KB and looks identical at normal viewing size. That is not a trick. It is what the format was built to do.

The catch is that this number is wildly different for different pictures. The same setting that is invisible on a portrait will wreck a screenshot of a spreadsheet, put grey fuzz around a logo, and turn a clear blue sky into a set of visible steps. The content of the image decides how much you can take.

What lossy compression actually discards

Lossy compression is not a uniform reduction in detail. It is a set of bets about what your eye is bad at, and it spends its budget accordingly. It keeps brightness detail and throws away colour detail, because human vision is far sharper for light and dark than it is for hue. It keeps broad shapes and throws away fine variation within small blocks of pixels.

That is why a photograph survives so well. Foliage, skin texture, fabric and sensor noise are all high-frequency mess, and losing some of it reads as normal. Your eye has nothing to compare it against.

It is also why the same compression is brutal on graphics. A black letter on a white background is a hard, high-contrast edge, exactly the kind of information the encoder is designed to smooth away. Nothing hides the damage, so you see it immediately.

Why 80 is the number, not 100 and not 60

The quality slider is not linear and it is not a percentage of anything meaningful. Going from 100 to 90 typically halves the file size and changes almost nothing you can see. Going from 90 to 80 saves another third for a difference you would need a side-by-side comparison to find. Going from 80 to 70 saves much less and starts to show.

That is the shape of the curve: enormous savings at the top, diminishing returns in the middle, visible damage at the bottom. Quality 80 sits at the elbow, which is why it has become the default in more or less every pipeline that has been tuned by someone paying attention.

Saving at 100 is close to pointless. It roughly doubles the file for detail no screen will show you, while still being lossy, so you get the cost of both approaches and the benefit of neither. If you genuinely need every pixel intact, use a lossless format instead of a maximum-quality lossy one.

Everything below 70 is a decision, not a default. Sometimes it is the right one, for a thumbnail nobody inspects or a background image sitting behind a dark overlay. Set it deliberately, and look at the result.

Every re-save takes another bite

Lossy compression is not idempotent. Open a JPEG, change one thing, save it again, and the encoder re-analyses an image that has already been damaged and damages it further. Do that ten times and the difference is obvious even to someone who has never thought about compression.

This is how image libraries decay over years without anyone making an obviously bad decision. A photo gets cropped, then someone rotates it, then it goes through an email client that helpfully re-compresses attachments, then it gets pulled back out and edited again.

The habit that prevents all of it: keep one lossless master, treat every published file as disposable output, and regenerate rather than re-edit. If a JPEG is all you have and it needs more work, move it to a lossless format first with JPG to PNG. That stops the bleeding, though it cannot recover what has already gone, and you can go back the other way with PNG to JPG when you export.

Screenshots and logos are a different problem

Save a screenshot as a JPEG and you will get ringing: faint grey ripples radiating out from every letter and every hard line, like a bad photocopy. It shows up at quality 90. At quality 80 it is unmistakable, and there is no setting that fixes it, because the artefact is the format working as designed.

Use PNG for these. Flat interface colour compresses brilliantly without any loss at all, so a screenshot is often smaller as a PNG than as a JPEG that also looks worse. The same holds for logos, charts, diagrams, code, pixel art and scanned line drawings.

The test takes two seconds. Zoom to 100 percent and look at the boundary between two colours. If it is a crisp step, you want lossless. If everything fades gradually into everything else, lossy is safe. Move between the two with the image format converter or PNG to WebP.

Resize first: it beats compression every time

This is the single most useful thing in this guide. Most images that are too big are too big in dimensions, not in quality. A 4000 pixel wide photo displayed in an 800 pixel column is carrying 25 times more pixels than anyone will ever see.

File size scales with area, so halving both dimensions cuts the pixel count to a quarter. That is a far bigger saving than dropping from quality 80 to quality 40, and unlike the quality drop it is completely invisible, because the extra detail was never being displayed.

So the order matters: resize with the image resizer, then compress with the image compressor. Doing it the other way round means you compress pixels you are about to discard, and you pay for the artefacts twice.

A practical rule for the web: no image needs to be more than twice its display width, to stay sharp on high-density screens. Beyond that you are shipping weight for nothing. If you are working out target dimensions for a layout, the aspect ratio calculator saves the arithmetic.

Four signs you have gone too far

Banding is the first to appear. Smooth gradients, skies and studio backdrops break into visible bands of flat colour with hard steps between them. It is the most common failure and the easiest to miss on a small preview, so check gradients at full size.

Blocking is the classic. Square patches of uniform colour, most visible in shadows and in flat areas, laid out on a regular grid. If you can see the grid, you are well past the point of no return.

Mushiness is subtler and more insidious. Fine texture stops being texture and becomes a smear: hair turns into a mass, grass turns into green paste, fabric loses its weave. The image still looks fine at thumbnail size and falls apart when anyone looks properly.

Halos and ringing around text and hard edges mean you have used a lossy format on something that needed a lossless one. Do not lower the quality further to fix it. Change format.

The short version

Start at quality 80 and only move if you have a reason. Resize before you compress, because dimensions dominate file size. Never put text or flat graphics through a lossy format. Keep a lossless master so repeated edits cannot grind the image down over time.

Then judge the result at the size it will actually be seen, on the parts of the image most likely to break. Compression is one of the few technical decisions where the correct answer really is a specific number, and that number is almost always 80.