Tools › Guides › What JPEG quality setting should you use?
What JPEG quality setting should you use?
What the number actually controls, why 100 is a waste, and how to find the point where a file stops getting better.
The number is not a percentage of anything
A JPEG quality setting of 80 does not mean "80% of the original quality". It is an index into how coarsely the encoder rounds off detail, and the scale is not linear, not standardised between programs, and not comparable across encoders. Quality 80 in one tool and 80 in another can produce noticeably different files.
What is consistent is the shape of the curve. File size falls steeply as you come down from 100, while visible quality holds almost flat, until it doesn't, and then it falls off a cliff.
Where to sit
| Setting | What it is for |
|---|---|
| 95-100 | Archival, or an image that will be edited again. Mostly wasted. |
| 85-92 | The sweet spot for photographs. Indistinguishable from the original at normal viewing size. |
| 75-85 | Web images where size matters more than pixel-peeping. Still good. |
| 60-75 | Thumbnails, backgrounds, anything small or behind text. |
| Below 60 | Visible blocking. Only for deliberate effect. |
If you want one number: 85. It is where almost all the savings are and almost none of the damage.
Why 100 is a waste
At quality 100 a JPEG is still lossy, it just rounds off almost nothing, and spends enormous numbers of bytes recording detail the format's own colour handling has already thrown away. A photograph saved at 100 is routinely three times the size of the same photograph at 92, for a difference invisible at any viewing distance.
If you genuinely need no loss, JPEG is the wrong format and PNG or WebP lossless is the right one.
The thing that actually costs you quality
Not the setting. The number of times the image has been through an encoder.
Every save re-compresses, and every re-compression works on an image that already has artefacts baked into it. Open a JPEG, crop it, save it, open it again, adjust it, save it again, and you have three generations of loss even if every save was at quality 95.
So: keep the original. Produce every output from it rather than from the last output. If you are making several sizes, make all of them from the master.
Target size versus target quality
Some tools let you say "make this 500 KB" instead of "use quality 80". That works by encoding repeatedly at different settings and converging on one that fits: usually about eight attempts, which is why it takes a moment longer.
It is the right mode when you have a hard limit to meet: an upload cap, an email attachment, a page budget. It is the wrong mode as a default, because it will happily destroy an image to hit a number you picked arbitrarily.
A well-behaved implementation stops when it is close enough rather than searching forever, and tells you what quality it landed on.
When JPEG is the wrong format entirely
JPEG was designed for photographs, smooth gradients, no hard edges. It is poor at exactly the things screenshots are made of:
- Text. Sharp black-on-white edges produce visible ringing, a ghostly
halo around every letter.
- Flat colour. Large areas of one colour get blotchy.
- Line art and logos. Same problem, more obvious.
- Transparency. JPEG has none at all. A transparent PNG saved as JPEG
gets a background, usually black or white, and you do not get to choose.
For any of those, PNG is smaller and better. For a screenshot of a web page containing both text and photographs, WebP handles the mix better than either.
What the file loses besides pixels
Re-encoding through a canvas, which is what any browser-based tool does, drops the EXIF block: camera, lens, timestamps and GPS coordinates. For a photo about to be posted publicly that is usually the point rather than a loss, but it is worth knowing before you compress your only copy.