What the Quality Slider Really Does, and How to Set It

The number is a coarseness setting, not a percentage

Lossy encoders break an image into small blocks and describe each block as a combination of frequency patterns: broad areas of flat colour are low frequency, fine texture and sharp edges are high frequency. Human vision is far more sensitive to the low frequencies. Compression exploits this by storing the low-frequency information precisely and the high-frequency information roughly.

The quality slider scales how roughly. A lower value means the encoder rounds more aggressively, so more of those fine-detail values collapse to zero and disappear from the file entirely. It is a dial on how much detail gets thrown away, not a percentage of anything you can measure.

This is why the same setting behaves so differently on different pictures. A portrait against a plain wall is mostly low frequency, so aggressive rounding barely shows. A photograph of gravel, foliage or a crowd is almost entirely high frequency, so the same setting visibly destroys it.

Where the useful range actually sits

The curve is steeply non-linear. Dropping from 100 to 90 often halves the file size with no perceptible change. Dropping from 90 to 80 might save another 30 percent. Going from 50 to 40 saves very little and costs a lot. Nearly all of the available saving lives between 100 and about 75, which is why that is where you should be working.

A repeatable method for finding the right setting

1. Pick three representative images from the set: your smoothest one, your most detailed one, and something typical. Do not calibrate on a single image.
2. Export the detailed one at 90, 80, 70 and 60.
3. View each at the size it will actually be displayed. Judging a 4000 pixel export at 100 percent zoom on a desktop monitor tells you nothing about how it looks at 600 pixels wide in a page.
4. Find the lowest setting where you cannot tell the difference from the 90, then add five as a safety margin.
5. Check that setting against the smooth image, which will tolerate more, and apply the result to the whole batch.

Recognising the three artefacts that matter

Blocking

Flat areas break into visible squares, usually eight by eight pixels. Most obvious in skies, shadows and out-of-focus backgrounds. It means the encoder no longer has enough precision to describe a smooth gradient, and it is the clearest signal that you have gone too low.

Ringing

A faint echo or halo alongside hard edges, particularly text over a flat background. This is the high-frequency information being approximated badly. It is the reason screenshots and diagrams should not be saved as lossy files at all.

Colour bleeding

Saturated colours, especially reds, smear past their boundaries. Encoders store colour information at lower resolution than brightness because the eye is less sensitive to it. At low quality settings this becomes visible on fine coloured detail.

Compressing a folder without ruining half of it

Batch processing applies one decision to many images, so the decision has to survive the worst case in the set, not the average one.

• Separate photographs from graphics before you start. They need different formats, not just different settings, and mixing them guarantees one group comes out wrong.
• Sort by content, not by folder. Product shots on white, outdoor photography and screenshots each want their own pass.
• Set the quality for the most detailed image in each group. The smooth ones will simply come out smaller than they needed to be, which costs you nothing.
• Spot-check the output, and specifically check the largest and smallest results. Outliers at either end usually indicate an image that belonged in a different group.
• Keep the originals. Compression is irreversible, and the setting that looked fine today will look wrong when the design changes and images get displayed twice as large.