Make an Image Low Quality Online

JPEG quality is the main control. It is the setting the browser’s encoder saves with, from 1 to 10 on the usual 1–100 scale, so even 10 is heavily compressed. At 1, most of every 8×8 block of pixels is reduced to a flat patch or a smooth gradient. Re-compressions saves and reopens the JPEG up to 20 times. Each pass uses your quality or up to three points above it, and the last uses exactly your setting, which is how a picture degrades when different apps save it in turn.

Resolution sets the size the work is done at, as a percentage of the original. With Stretch back to original size switched on, the small version is enlarged again with nearest-neighbor scaling, so every remaining pixel becomes a square block: the pixelated, low-res look. Switch it off to download the small image itself.

Blur softens the picture before it is compressed, and Noise makes it grainy, like fast film. Colors per channel cuts red, green and blue to a few values each, an effect also called bit crushing; six values per channel give the 216-color palette early web designers relied on. Color bleed smears color sideways while brightness stays put, much like a worn videotape.

The download keeps the original’s proportions. Transparent areas turn white, because every pass is a JPEG and JPEG has no transparency. Animated GIFs open at their first frame; to work on another one, take it out with the GIF splitter, and to put degraded frames back together, use the GIF maker.

A JPEG does not store pixels one by one. The encoder first separates brightness from color, and most encoders then keep the two color channels at half the width and half the height, because our eyes notice fine color detail far less than fine brightness detail. This is chroma subsampling, and it is why strong reds and blues look smudged along their edges in a heavily compressed picture.

Each channel is then cut into 8×8 blocks, and each block is described as a mix of 64 standard patterns, from completely flat to finely striped (the discrete cosine transform). Quantization divides those amounts by numbers from a table and rounds the results, and the quality setting scales the table. At high quality the rounding is gentle. At low quality the fine patterns round to zero, each block keeps little more than its average color and a slope, and the grid of squares becomes visible, especially in skies, walls and skin.

Ringing, sometimes called mosquito noise, is the other giveaway. A sharp edge such as black text on white needs exactly the fine patterns that low quality discards, so the edge is rebuilt with ripples and faint dots around it.

Generation loss is what happens when a JPEG is saved again. Each save quantizes an image that already contains artifacts, and some of them grow. Saved again with exactly the same settings, though, a JPEG barely changes after the first save, because its blocks already hold values the encoder reproduces. The loss returns whenever something changes between saves: a crop, a resize, or another app with a different quality setting. That is why pictures that are screenshotted and reposted many times end up so badly degraded, and why the re-compressions here vary the setting slightly from pass to pass.

To make a deep-fried image, switch the effect to Deep fry and pick a preset. Deep fried raises saturation to 230% and contrast to 150%, sharpens hard, adds grain and a red-orange cast, then saves the result eight times at JPEG quality 4. Nuclear goes further and puts a bulge in the middle of the picture; click anywhere on it to move the bulge, or change Bulge size to cover more of it.

Deep-fried memes are images edited until they look burnt: colors pushed far past natural, harsh edges, grain and obvious compression. The usual recipe is the one the presets follow, with saturation and contrast up, heavy sharpening and noise, then low-quality JPEG saves. The order matters. Sharpening and noise added before compression give the encoder more fine detail to mangle, and that is where the crunchy texture comes from.

According to Know Your Meme, the earliest known post to use the name “deep fried meme” appeared on Tumblr in March 2015, although memes were being edited this heavily before then. This tool does not add text or emoji; to caption the result, open the downloaded file in the Photo Editor.

Lowering quality on purpose usually makes the file smaller too, but that is a side effect. A compressor works the other way round: it keeps the picture looking as good as possible and finds the highest quality that fits. To email a photo or get under an upload limit with no visible damage, use the image compressor, or shrink the image to a target size such as 500 KB.

A result from this tool can even be larger than the original. Stretching a 160 px version back to 4,000 px and saving it as PNG stores every enlarged block, and that takes space. The size shown under the preview is the size of the file you will download, so you see it before saving. To check the dimensions and file size of any picture later, use the image size checker.

A blurry, blocky picture can still show a face, a name or a number plate well enough to identify it, and people can recognize someone they know even in a very small, blurry photo. Do not rely on this tool to hide anything. To cover faces, text or plates, use the image pixelator with large blocks or a black box, or crop the area out completely.

The picture itself never leaves your device: it is opened, degraded and saved by your browser. The downloaded file is a new image, so the original’s EXIF data, including any GPS location, is not copied into it.