Compress an image without losing quality
Free savings come from removing metadata and choosing the right format. Invisible savings come from resizing to the size the viewer actually needs. Everything beyond that is lossy, and the practical goal is staying above the threshold where JPEG artefacts become visible.
Open the Compress to Size tool
"Without losing quality" is true for part of the job and false for the rest. Knowing which part you are in tells you how far you can push.
Category one: free
- Metadata. GPS coordinates, camera settings, embedded thumbnails. Removing them changes no pixel and can save dozens of kilobytes.
- Right format. A screenshot stored as JPEG is both larger and uglier than the same screenshot as PNG. A photograph stored as PNG is many times larger than the JPEG. Choosing correctly costs nothing.
- Lossless re-encoding. Some encoders store the same pixels in fewer bytes by removing redundant data. Bit-identical output, smaller file.
Category two: invisible
Resizing to what the destination needs. If an image is displayed at 800 pixels wide, storing 4000 pixels is waste that no screen can reveal. This is technically lossy — those pixels are gone — but there is no visible difference at any realistic viewing size.
The caveat is future use. A 4000-pixel original can be printed, cropped, or shown on a large display later. That is why you keep the original and send the resized copy.
Category three: damage
JPEG quality reduction. The scale is 0–100 and the useful range is narrower than it appears:
| Quality | What happens |
|---|---|
| 90–100 | Visually identical to the source; large files |
| 80–90 | The practical default — no visible difference at normal size |
| 70–80 | Slight softening in fine detail; usually acceptable |
| 50–70 | Visible blockiness around edges and text |
| Below 50 | Clearly damaged; colour banding appears in gradients |
The habit that does the most damage
Repeated compression. Saving a JPEG, then compressing that JPEG again, re-encodes the already-damaged data. Each generation accumulates artefacts, and the second pass usually does more visible harm than the first. If a file is still too large, go back to the original and target lower in one pass.
The same applies to screenshots of compressed images, images downloaded from chat apps, and anything that has been through three tools already.
A workflow that holds up
- Keep one untouched original, always.
- Strip metadata and pick the right format.
- Resize to the destination's real display size.
- Set the byte target in one pass, quality no lower than needed.
- Check the busiest area of the image at 100%.
When the limit forces damage
20 KB for a detailed photograph is below the threshold where quality survives. At that point the honest options are to reduce the subject matter — crop tighter — or to accept that the form's limit was designed for headshots, not landscapes. No encoder can put back detail the budget cannot afford.
Questions
Is any image compression truly lossless?
Yes, when it only removes metadata or re-packs identical pixel data. Shrinking a photograph's actual content is always lossy.
What JPEG quality is visually lossless?
Around 85–90 for most photographs. Below 70 artefacts become noticeable on text and sharp edges.
Why does my image degrade after several saves?
JPEG compression is generational. Each save re-encodes the previous result, and the damage accumulates.
Does WebP lose less quality at the same size?
At a fixed byte budget, WebP usually looks better than JPEG. If the destination accepts it, it is the better choice.
Updated 2026-11-16 · written for this site, no trackers, no account.