Screenshots and Photos Need Opposite Treatment
There are two fundamentally different kinds of image, and almost every piece of advice about formats and compression applies to one of them and actively harms the other. Once you can tell which you are holding, most image quality problems answer themselves.
Continuous Tone Versus Synthetic
A photograph is what engineers call continuous-tone. Brightness and colour vary smoothly and unpredictably across the frame. Adjacent pixels differ slightly. There is sensor noise everywhere, even in areas that look uniform. Edges are soft, because lenses and the physical world produce gradual transitions.
A screenshot is synthetic. It was drawn by software onto a pixel grid. Large regions are exactly one colour, repeated perfectly. Edges are absolutely hard — a letter is black, the pixel beside it is white, with nothing in between. There is no noise at all, because no sensor was involved.
These are not slight variations on a theme. From a compression standpoint they are opposite problems, and the tools that excel at one are exactly the wrong choice for the other.
Why JPEG Destroys Text
JPEG works by converting each small block of the image into frequency information and discarding the high frequencies — the finest, sharpest detail — on the reasonable assumption that your eye will not miss them.
For a photograph, that assumption holds. The discarded high-frequency content is largely noise and micro-texture. For text, the assumption collapses completely, because a hard black-to-white edge is almost entirely high-frequency information. It is precisely what JPEG is designed to throw away.
The result is the coloured haze you have seen around lettering in a compressed screenshot. Engineers call it ringing; it is the encoder trying to reconstruct an infinitely sharp edge from frequency components it no longer has. Worse, because a screenshot is mostly flat colour, JPEG does not even save much space in exchange. You get a visibly degraded image at roughly the same file size.
If you zoom into text in a compressed screenshot and see faint colour smudging around the letters — often reddish or greenish where the original was pure black and white — that is JPEG ringing. No amount of sharpening will remove it, because it is not blur; it is fabricated data replacing information that was discarded.
Why PNG Is Wrong for Photographs
The reverse mistake is less visible but more expensive. PNG is lossless: it must reproduce every pixel exactly, and it achieves compression only by finding repetition.
A screenshot is full of repetition, so PNG compresses it brilliantly. A photograph has essentially none — every pixel of sky is a slightly different blue, and sensor noise ensures even flat surfaces are not truly flat. PNG dutifully stores all of it, producing files five to ten times larger than a JPEG that looks identical.
Nothing looks wrong, which is why this mistake persists. The cost is paid in slow-loading pages, bloated email attachments, and storage consumed for no perceptible benefit.
| Photograph | Screenshot | |
|---|---|---|
| Content | Continuous tone, noisy | Flat colour, hard edges |
| Best format | JPG or WebP lossy | PNG or WebP lossless |
| Worst format | PNG — huge for no gain | JPG — ringing around text |
| Compresses because | Detail can be discarded | Patterns repeat |
| Resizing | Downscale freely | Avoid — text degrades fast |
| Quality setting | 75 to 90 is plenty | Use lossless instead |
Resizing Splits the Same Way
The divide does not stop at format choice. Downscaling a photograph is close to free, and often improves it — averaging groups of pixels together suppresses sensor noise, so a well-downscaled photo can look crisper than the original.
Downscaling a screenshot damages it immediately. Text was rendered to align with the pixel grid; the stems of letters are frequently exactly one pixel wide. Resample that to 80 percent and every stem now falls between pixels, so the resampler averages it into a grey smear. The text becomes fuzzy at a size where it should still be perfectly legible.
If a screenshot must be smaller, the better approach is usually to crop it rather than scale it — show less of the screen at full sharpness instead of all of it blurred. Where scaling is unavoidable, exact halves work considerably better than arbitrary percentages, because each output pixel maps cleanly onto a whole number of input pixels.
The Hybrid Case, and What to Do About It
Real work is not always tidy. A social media graphic is a photograph with text laid over it. A product shot has a logo in the corner. A slide has a chart on a photographic background.
These are genuinely difficult, because the image contains both kinds of content and any single setting compromises one of them. JPEG will smear the text; PNG will store the photograph inefficiently.
WebP handles this better than either, which is one of the strongest practical arguments for using it. Its lossy encoder deals with sharp edges considerably more gracefully than JPEG, so text survives at settings where JPEG would visibly damage it, while photographic areas still compress well.
Where WebP is not an option, decide which content matters more. If the text carries the message — a headline, a price, a call to action — protect it and accept a larger file. If the photograph is the point and the text is decoration, compress normally and let it soften.
A Practical Checklist
- Did it come from a camera? Use JPG or lossy WebP. Quality 75 to 90 is ample.
- Did it come from a screen? Use PNG or lossless WebP. Do not run it through a JPEG encoder.
- Does it contain readable text? Never compress it heavily, whatever it is.
- Does it need transparency? PNG or WebP. JPG cannot store it at all.
- Is it going on a website? WebP, with a JPG or PNG fallback for anything that leaves the browser.
- Is it a mixed graphic? Prefer WebP, and protect whichever element carries the meaning.
Almost every image quality complaint traces back to violating one of these. Most often it is a screenshot that went through JPEG somewhere it should not have, or a photograph sitting on a web page as a multi-megabyte PNG.
Frequently Asked Questions
What format should I use for screenshots?
PNG, or lossless WebP if the destination supports it. Both store every pixel exactly, so text stays perfectly sharp, and both compress the large flat areas of a screenshot very efficiently. Avoid JPEG entirely — it produces visible colour fringing around lettering while barely reducing the file size.
Why do my screenshots look blurry after uploading?
Usually one of two things. Either the platform re-encoded the image as JPEG, which smears hard edges, or the screenshot was scaled to fit a layout, which pushes one-pixel-wide letter stems between pixels and averages them into grey. Uploading a PNG at exactly the display size avoids both.
Can I fix a screenshot that has already been through JPEG?
Not really. The ringing artifacts around text are fabricated data that replaced the original information, so there is nothing to recover. Sharpening makes it worse by emphasising the artifacts. The only genuine fix is to retake the screenshot and save it as PNG this time.
Is WebP good for both photographs and screenshots?
Yes, which is unusual. It has separate lossy and lossless modes, so it handles continuous-tone photographs and flat synthetic graphics well, and its lossy encoder deals with sharp edges far better than JPEG. That makes it the best single choice for mixed content, provided the destination is a browser.
Should I resize screenshots down to save space?
Prefer cropping. Text is rendered to align with the pixel grid, so scaling it to an arbitrary size blurs the letterforms almost immediately. Cropping to show only the relevant part keeps everything perfectly sharp. If you must scale, exact halves work far better than arbitrary percentages.