English ▾
Free, Fast & Secure Online Tools
It feels backwards. A photograph contains a whole scene — thousands of colours, foliage, texture, noise — and arrives at 800 KB. A screenshot of a mostly white page with some text on it arrives at 3 MB. The simpler picture is four times the size.
Three separate things cause this, and they multiply rather than add.
Nearly every system saves screenshots as PNG, which is lossless — it reproduces every pixel exactly, discarding nothing.
JPEG, which your camera uses, is lossy. It throws away detail the eye is unlikely to miss, and that is where its enormous size advantage comes from. A photograph is forgiving of this: its content is already irregular, so approximations hide easily.
So the comparison is not really “simple picture versus complex picture”. It is “nothing discarded versus most of it discarded”.
This is usually the largest factor and the one people are least aware of.
On a high-density display — any recent phone, most laptops — the screen reports a convenient size while containing far more physical pixels. A laptop that behaves like a 1440-pixel-wide screen may have 2880 actual pixels across.
A screenshot captures the physical pixels. So a capture of a window that looks 1000 pixels wide produces a 2000-pixel-wide image: four times the pixel count of what you thought you were grabbing. Open a screenshot and check its dimensions — the number is usually a surprise.
Recent Apple hardware and many newer phones capture in Display P3 rather than sRGB, and the screenshot carries a colour profile to say so.
The size effect is modest compared to the first two. The reason to know about it is different: it is why a screenshot can look oddly oversaturated when someone else opens it in software that ignores the profile.
This is the obvious fix and it is usually wrong.
JPEG compresses by approximating, and it handles hard edges badly. Text, window borders, icons and UI lines are nothing but hard edges. Save a screenshot of text as JPEG and you get a grey halo around every letter and a faint ringing along every border. Small text becomes noticeably harder to read.
PNG is genuinely the correct format for a screenshot of an interface. The problem is not the format — it is that nothing has been done to the image.
In rough order of how much they save:
Crop to what matters. Most screenshots are a full screen when the point is one dialogue box. Cropping to the relevant quarter removes three quarters of the data, costs nothing in quality, and makes the screenshot clearer to the person receiving it. This is almost always the biggest win available.
Reduce the colour count. A photograph may contain a hundred thousand distinct colours. A screenshot of an application window might contain two hundred. PNG can be saved with a 256-colour palette instead of full colour, which often cuts the file by 60–80% with no visible difference at all on interface content. Many compression tools do this automatically; it is the single most effective step for screenshots specifically.
Scale down to 1×. If the screenshot will be viewed at normal size rather than examined closely, halving the dimensions removes three quarters of the pixels. Keep the full resolution when fine detail matters — a bug report where a one-pixel misalignment is the point.
Use WebP instead. Lossless WebP typically produces files around 25% smaller than PNG with identical pixels. If the destination accepts it, it is free savings.
Combining the first two usually turns a 3 MB screenshot into something under 200 KB that looks exactly the same.
If what you captured is mostly a photograph — a web page dominated by a hero image, a video frame — the logic flips. That content is photographic, PNG is a poor fit for it, and JPEG or lossy WebP will be dramatically smaller with no meaningful loss.
The rule is about content, not about the fact that it came from a screen: hard edges and flat colour want PNG; continuous tone wants JPEG or WebP. A screenshot that mixes both is a judgement call, and lossy WebP handles the mixture better than JPEG does.
A lot of oversized screenshots exist because someone screenshotted text instead of copying it.
An error message pasted as text is perhaps 200 bytes. The same message as a screenshot is two megabytes — ten thousand times larger — and it cannot be searched, selected, translated, read aloud by a screen reader, or copied into a search box by the person trying to help you.
Screenshot the interface when the layout is the point. Copy the text when the text is the point.