Meme Generator: Add Top and Bottom Text to Any Image in Your Browser

By Iain · Melbourne VIC, Australia

A meme generator does one job: it puts text on a picture and hands the picture back. That job is small enough to run inside a browser tab, and doing it there removes most of the reasons people abandon online editors halfway through a caption. The generator on this page opens with a drop zone, accepts an image of up to 15 MB, and places two caption boxes on the canvas before you type a single word. Everything after that is text placement, type sizing, and export.

This article covers what the tool does, why the classic caption style looks the way it does, which settings decide whether a caption survives being viewed on a phone, and how image dimensions interact with the platform you plan to post to.

What the generator does with your image

You drop in a PNG, JPG, WebP, or GIF, and the file is read locally as a data URL. Nothing is uploaded to a server, which matters if the picture is a screenshot of an internal dashboard, a photo of a friend, or anything else you would rather not hand to a stranger's storage bucket. The image is drawn onto a canvas at its natural pixel dimensions, so a 1200 by 630 screenshot stays 1200 by 630 and an 800 by 800 template stays square.

Two caption boxes appear by default. The first sits at the top of the image, four per cent of the height from the upper edge. The second sits near the bottom, eighty-four per cent of the way down, which keeps it above the very edge rather than flush against it. Both start centred horizontally, both use bold Impact at 48 pixels with a four-pixel black outline, and both carry placeholder text you replace immediately.

Above the canvas you can add as many further text boxes as the joke needs. Each box can be dragged to any position on the image, which is how the tool handles formats that are not the classic two-line template: a caption over a face, a label over an object, a third line on a wide screenshot. Each box keeps its own font size, outline width, colour, and alignment, so a single image can mix a large headline with a small annotation.

When you are finished, the canvas is exported as a PNG at the image's native resolution and saved to your device with the source filename plus a suffix, so a file called cat.jpg produces cat_meme.png. There is no watermark, no account, and no export queue.

Why the classic caption uses Impact with a black outline

The familiar treatment, white capital letters with a heavy black edge, was not a style choice so much as a solution to a layout problem. Captions sit on top of photographs whose colours and brightness you cannot control: a sunlit beach, a dark room, a busy cartoon still. A single flat text colour reads clearly against some backgrounds and disappears against others, and there is no colour that stays legible over every possible image.

An outline fixes that structurally. The black stroke separates the letters from whatever is behind them, so the white fill always has something dark to sit against and the letters keep their shape. Two to three pixels of stroke is the historic figure, and it works because the edge is thick enough to hold at thumbnail size without swallowing the letterforms.

Impact itself contributes three properties. Its strokes are heavy and uniform, so they survive compression and downscaling. Its letterforms carry little internal detail, so nothing blurs into the counters when the image is resized for a feed. Its spacing is tight and its shapes are condensed, which fits more characters into a given width, a requirement of the two-line top-and-bottom format where each caption is a single line if the writer is being disciplined.

The style has a limit. It is a reliable default over unpredictable backgrounds, not a rule, and captions designed for a specific layout often read better in a different face. What matters is the underlying principle: heavy weight, minimal detail, and contrast between the text and the pixels immediately behind it.

Placing a caption without guessing

The default positions, four per cent from the top and eighty-four per cent from the bottom, exist because captions near the edges stay out of the way of faces and focal points in the middle of the frame. They also sit inside the area most platforms leave alone when they crop or overlay interface elements.

Dragging a box is the adjustment most people reach for first, and it is usually the right one when the subject of the photo sits off-centre and the caption would otherwise cover it. A long caption wraps automatically rather than running off the canvas, and the line spacing is calculated as 1.1 times the font size, which keeps consecutive lines from touching without introducing the wide gaps that make a two-line caption look like two separate comments.

Alignment becomes relevant when a caption is longer than one line. Centred text is the convention for the classic format because it balances the two edges of the block. Left alignment suits a caption placed beside an object or over a screenshot where the text behaves like a label, and right alignment does the same from the opposite side.

The rule that keeps captions readable is length. A line of text that wraps to four or five lines at full width stops being a caption and starts being a paragraph, and the joke it carries gets slower to read at exactly the moment it needs to be fast.

Type size, outline width, and contrast

Font size in the generator runs from 12 to 160 pixels and starts at 48. General guidance for meme captions puts the comfortable band between 48 and 72 pixels for standard image sizes, with roughly 52 to 60 pixels being the point where text stays large enough to read in a phone feed without dominating the picture. Shorter captions can be pushed larger, and longer ones need to come down a step or two, because a caption that fills the width at 72 pixels will not fit at all once it becomes two lines.

Outline width runs from 0 to 20 pixels and starts at 4. The value does the same job the letterform does: it buys contrast. Raising it beyond about six pixels on small text starts to close the gaps inside letters such as e and a, so the practical range for a caption at 48 pixels is between two and five pixels of stroke. Drop the outline to zero and a white caption over a bright sky becomes unreadable, which is the single most common mistake in hand-made memes.

Text colour defaults to white and alignment defaults to centre. White over a dark outline is the combination that works across the widest range of photographs. Changing the fill colour is worthwhile when the image has a light background and the caption can be set in a dark tone, which keeps the visual weight of the text separate from the picture instead of fighting it.

Choosing image dimensions for the platform

The aspect ratio of the source image decides more about how a meme travels than any typography setting, because platforms crop, letterbox, or compress whatever does not match their expected shape. Composing on an image that already has the target ratio avoids the crop that would otherwise cut the caption off.

Classic blank templates cluster around 500 to 800 pixels square, which is enough for a two-line caption but well below what modern feeds prefer. Starting from a larger source and staying under roughly 3 MB gives the compression algorithms something to work with, since most services visibly degrade images above about 5 MB.

There is a measurable return on this step. A 2025 study by Hootsuite found that posts carrying correctly sized images received 23 per cent higher engagement than posts with images that had been wrongly cropped or uploaded at low resolution, and the work involved is a single decision made before the caption is written.

Export quality and file size

The generator writes a PNG at the image's natural resolution, which means the output has the same pixel dimensions as the input and no resampling has taken place. That is the right default for memes built from screenshots and templates, because PNG stores text edges without the ringing artefacts that a lossy format introduces around high-contrast letters.

The trade-off is size. PNG files are larger than JPEG for photographic sources. A 1080 by 1080 PNG exported from a photograph can run several megabytes, while the same image as a JPEG would be a fraction of that. Since most platforms re-encode uploads anyway, a large PNG is not usually a problem for quality, but it can matter for chat apps with tight attachment limits.

The input limit is 15 MB. Photographs straight from a modern phone camera sit below that comfortably, while very large scans and multi-frame GIFs can exceed it. Reducing the source image in an image resizer before loading it here is faster than working around the limit afterwards.

Classic two-line format versus caption above the image

The two-line layout is not the only convention in circulation. A caption placed above the picture on a clean background, in ordinary sentence case rather than capitalised Impact, has become the dominant modern shape, largely because screenshots of social posts are themselves a meme format.

Imgflip's own figures put 67 per cent of new meme creations in 2025 in the caption-above-image style rather than text over the picture, which is a significant shift from the period between 2012 and 2015 when the Impact caption peaked. The older treatment remains the correct choice for established templates, where the visual grammar of the format is part of the joke.

Template usage concentrates heavily. Imgflip data shows the Drake Hotline Bling comparison format has been the most-used template for five consecutive years, with more than five million generations, and research cited by Know Your Meme indicates comparison formats account for about 38 per cent of viral memes since 2020. Their structure does half the comedic work, which is why the format outlives the specific images it was built from.

A working method that takes two minutes

1. Decide where the meme will be posted and pick a source image with that aspect ratio.
2. Load the image, then replace the two default captions before touching positioning.
3. Set the font size so each caption is one line, or two short lines at most.
4. Drag the boxes clear of faces, then check the stroke width against the busiest part of the background.
5. Export the PNG and confirm the file size sits comfortably inside the destination's limit.

Watermarks undermine the last step. An analysis of meme-focused subreddits in 2024 found that posts carrying a visible generator watermark received 41 per cent fewer upvotes on average than clean copies of the same content, which is a large penalty for a mark nobody asked for. The generator on this page adds none.

Iain

Melbourne VIC, Australia · SO reputation 2000 · Badges: 2🥇23🥈40🥉 · SO member since 2009