What the slider is doing
In a lossy format, the quality setting decides how much detail the encoder is allowed to discard. High values keep almost everything and produce larger files; low values throw away more, and the encoder spends its remaining budget on the parts of the image where detail is most visible. The relationship is not linear: dropping from 95 to 85 usually saves a lot for very little visible cost, while dropping from 60 to 50 saves less and costs noticeably more.
Lossless formats have no such dial. PNG either stores the pixels exactly or does not exist — its “level” only controls how efficiently the identical data is packed, which changes the file size but never the image.
Sensible starting points
- **Photograph, general use — quality 82–88.** Usually indistinguishable from the original on screen and comfortably smaller.
- **Photograph with a hard byte limit — quality 70–80, then adjust dimensions.** Below about 70, banding and blocking tend to appear in smooth areas such as skies or skin.
- **Screenshot or scanned text — quality 88–95, or PNG.** Text is high-frequency detail and shows artefacts early; if size must drop, reduce dimensions instead of quality.
- **Thumbnail or small preview — quality 65–75 is fine.** At small display sizes, artefacts are far less visible.
- **Archive copy you may need later — do not compress at all.** Keep the original somewhere and compress a working copy.
These are starting points, not rules. A noisy, detailed photograph and a plain studio shot behave very differently at the same setting — always judge by looking at your own image.
How to judge “too far”
- Open the compressed result and view it at 100% zoom — not fit to screen, which hides artefacts behind scaling.
- Look at three places: smooth gradients (banding and blocks), hard edges around text or logos (ringing halos), and fine texture such as hair or fabric (smearing).
- Compare against the original side by side if you can keep both open.
- If any of those three is clearly worse, raise the quality one step rather than restarting.
- Only compress once from the original. Every additional save cycle is another generation of loss.
Suppose a photo at 3000 × 4000 must be under 100 KB and quality 70 still gives 240 KB. Pushing quality down to 40 might reach the target but the face would show heavy blocking. Reducing to 1200 × 1600 first then using quality 75 usually lands near the target with a photo that still looks clean — because four sixteenths of the data was removed by the resize, leaving the quality setting free to stay high. The exact numbers depend on the image; the trade-off does not.
Compression in the tools on this site
- Compress Image gives you the direct dial with a live before/after size readout.
- Exact KB Photo Compressor inverts the problem: you give it a target and a tolerance and it finds the setting.
- Image Converter applies a chosen quality while changing format, which is often the bigger win than the quality change itself.
- Assignment Compressor applies one consistent maximum edge and quality across a whole hand-in so a batch comes out uniform.
Common questions
Is a lower quality number always a smaller file?
For the same image and format, yes. Across different images or formats the numbers are not comparable — quality 70 on a flat image can be smaller than quality 85 on a noisy one.
Should I compress once or several times?
Once, from the original. Each re-encode of an already-lossy file discards a further layer of detail and the losses accumulate.
Does compression change my image's dimensions?
No. Pure compression changes encoding only. If dimensions must change, do that deliberately as a separate step so you can verify both values independently.