How to Reduce Image File Size Without Unnecessary Quality Loss
Shrinking a JPEG always costs something. The goal is to lose the least that gets you under the limit: reduce only as much as needed, and choose the right lever.
“Without losing quality” needs a footnote
A JPEG is already a lossy format. Making one smaller means discarding more detail, so some loss is unavoidable. What you can control is how much you lose and where. A blind approach (drag a slider until the file is small enough, then re-save) usually loses more than needed. The better approach is to ask for a target and let the tool find the highest quality that meets it.
The three levers
- Pixel dimensions. Fewer pixels means a smaller file, and often the best trade. A phone photo is far bigger than any screen shows. Scaling it to a sensible size removes bytes the viewer would never see. See image dimensions vs file size.
- Compression quality. The JPEG encoder’s setting. Lower means smaller and blockier. The effect is gentle near the top and steep near the bottom.
- Format. A photo as PNG is much larger than the same photo as JPEG, because PNG keeps every pixel. A screenshot of text can go the other way. See PNG vs JPG for online forms.
In Compress Image you can pick a quality yourself and watch the resulting size, or give a target size and let it search. Target mode keeps the highest quality that fits and only shrinks pixels if that isn’t enough.
What the numbers mean
A quality-aware tool shows several numbers. They answer different questions, and mixing them up is how people end up trusting the wrong one.
| Number | What it is | What it is not |
|---|---|---|
| Compression quality 85 / 100 | The encoder setting that produced the file | Not “85% of the quality kept” |
| 47% smaller | A plain comparison of file sizes in bytes | Not a measure of how the picture looks |
| Visual similarity 99% | A measured structural similarity (SSIM on brightness) between original and result | Not a guarantee, and shown only when the result has the same shape as the original |
On the third one, the comparison is made on a copy scaled to at most 1024 px on the long side. That scaling hides fine noise, so for large photos the score is somewhat optimistic. Treat it as a sanity check, and look at the picture.
Guardrails
When you give Compress Image a size, it works within default floors rather than ruining the picture: JPEG quality 50, at least 55% of the size on each side, and a 400 px minimum on the long edge. If the limit is out of reach above them, it tries harder compression once and keeps that only if the result still measures at least 90% similar to the original. Failing that, it returns the best result at the floor and says “target too aggressive”. It also warns when quality falls below 60/100. And if your file is already within the limit, it is returned untouched.
A short checklist
- Decide whether the pixel size can come down. It usually can.
- Prefer JPEG for photographs.
- Set a target, not a guess, and read the quality it landed at.
- Zoom in on faces and text before you submit.