Split a long screenshot

A scrolling capture is often too tall to send, print or post. Cut it into pieces here, with the cuts landing in the gaps rather than through a line of text. It all happens in this browser tab and nothing is uploaded.

How it works

A scrolling capture solves one problem and creates another. The whole conversation, the whole receipt, the whole article is now in a single image, and that image is 1,170 pixels wide and 14,000 tall. Messaging apps squash it into an unreadable ribbon, printers give it one page and a wall of white, and posting it anywhere means it arrives as a thumbnail nobody can read.

Cutting it up is the fix, and the interesting part is where to cut. Divide the height evenly and the cuts land wherever the arithmetic puts them, which is as likely to be halfway down a row of characters as anywhere else. Half a sentence at the bottom of one image and half at the top of the next is what makes a split capture look broken, and it is entirely avoidable.

So each cut is treated as a suggestion. Before anything is drawn, the rows within about a line and a half of every cut are scored by how often the colour changes along them: a row crossing text changes at every character edge, a row in the gap between paragraphs does not change at all. The cut moves to the quietest row nearby, which in a screenshot is the space above a message bubble or below a heading. Pieces come out very slightly uneven, which nobody notices, instead of slicing a word in half, which everybody does.

Only the rows a cut could actually reach are examined. Reading the whole capture would be the obvious implementation and the wrong one: a 1,000 by 20,000 capture is 80 MB of raw pixels, enough to crash a tab on a phone, and all but a few hundred of those rows could never have been chosen. Square tiles are the exception to the whole scheme, because there a moved cut is a tile that is no longer square, which was the entire request; those cut exactly and let the last tile be short.

How to use it

1 Choose a tall screenshot, or paste one straight from your clipboard.
2 Say what you need: screen-sized pieces, a set number of them, or square tiles.
3 Press Split. The cuts are nudged into the gaps between lines as it works.
4 Download the pieces, as one image or as a ZIP.

Good to know

An even division beats slicing at a fixed height, which leaves a useless sliver at the end.
Pieces come out as PNG, because a screenshot is text and flat colour, which is what PNG keeps sharp and JPEG smears.
Files are numbered with a leading zero where needed, so they sort correctly in a file manager.
Square tiles are cut exactly, so they stay square for a carousel; every other mode nudges its cuts.
If you need the opposite, the screenshot stitcher joins several captures into one.

Browser support

Chrome 94 and later
Firefox 93 and later
Safari 16.4 and later
Edge 94 and later

These are the oldest versions with everything the tool uses. They are worked out from the tool itself, so they change when it does.

Version numbers checked on 2026-09-04.

Where to cut a long screenshot, and why the obvious answer is wrong

Splitting an image into pieces is arithmetic that a beginner can write: take the height, divide by the number of parts, cut there. It works perfectly and produces a bad result, because the one thing the arithmetic cannot know is that row 4,000 happens to run straight through the middle of a line of text. Half a sentence at the bottom of one image and half at the top of the next is the signature of a split done this way, and it is the reason people go back to doing it by hand in an image editor.

The fix is to stop treating a cut as a position and start treating it as a preference. The even division says where the cut would ideally go; the picture gets a say in where it actually goes. Allow each cut to move by around a line and a half, and there is almost always somewhere better within reach, because text is not continuous. It comes in lines with gaps between them, in bubbles with padding around them, in sections with rules between them. A capture that looks dense is mostly empty rows.

Finding them needs a measure of how busy a row is, and the cheapest useful one is counting transitions: walk the row and count how often the colour changes by more than a hair. A row crossing text changes at every character edge and scores in the hundreds. A row in the gap between paragraphs scores zero. This is not edge detection, and deliberately so: it runs on a phone over tens of thousands of rows, and one comparison per pixel with no arithmetic is what keeps it instant.

Two rules keep the nudging honest. A cut only moves somewhere strictly quieter than where it already is, so a photograph or a gradient, which has no quiet rows anywhere, keeps its even division instead of being pushed around by noise. And among equally quiet rows the nearest wins, because a run of thirty blank rows would otherwise always snap to its first one and drag the cut further than it needed to go.

The last decision is about memory, and it is the one that separates a demo from a tool. Scoring every row means reading every pixel, and a 1,000 by 20,000 capture is 80 MB of RGBA, which is enough to end the tab on a phone, and a phone is exactly the device the capture came from. But a cut can only move as far as its radius, so all but a few hundred of those rows were never candidates. Reading one narrow band around each cut costs a fraction of a megabyte and answers precisely the same question. The general lesson is worth keeping: when an algorithm looks too expensive, check whether it is being asked about data whose answer could never have mattered.

Frequently asked questions

How do I split a long screenshot into parts?

Choose the image, pick what you need, and press Split. Screen-sized pieces suit sending and reading, a set number suits a fixed layout, and square tiles suit a carousel post. Several pieces arrive together as a ZIP.

Will it cut through the middle of a sentence?

It tries hard not to. Each cut can move by around a line and a half to reach the quietest row nearby, which in practice is the gap between paragraphs or above a message bubble. On a photograph or a gradient there are no quiet rows, so the even division is kept rather than nudged around by noise.

Can I split an image into equal parts?

Yes. Choose a set number and the height is divided evenly between them. The cuts still shift slightly to miss the text, so the pieces are equal to within a few pixels rather than exactly.

Can I make an Instagram carousel from a tall image?

Choose square tiles. Each tile is as wide as the image and just as tall, so it is genuinely square, and the cuts are left exactly where the arithmetic puts them because moving one would stop the tile being square. The final tile may be shorter than the rest.

Does splitting lose quality?

No. Each piece is drawn from the original pixels and saved as PNG, which is lossless. A screenshot is mostly text and flat colour, so PNG is both the sharper and usually the smaller choice.

Is my screenshot uploaded?

No. It is decoded, cut and re-encoded inside this browser tab by your own device. You can keep your browser’s Network tab open while it works: nothing carrying the image goes out.

Your files are processed on your device and are not uploaded. You can verify this in your browser’s developer tools under the Network tab.