What image compression actually means
When you compress an image, you're asking an algorithm to store the same visual information using less disk space. The core question is: how much information can you remove before the human eye notices?
The answer depends entirely on the format. JPEG, PNG, and WebP use fundamentally different mechanisms. Understanding these differences lets you choose the right settings every time instead of guessing.
JPEG: perceptual block compression
JPEG exploits a key property of human vision: our eyes are far more sensitive to changes in brightness than changes in color, and far more attentive to broad areas than fine peripheral details.
The algorithm divides the image into 8×8 pixel blocks and applies a mathematical transform called DCT (Discrete Cosine Transform) to each. This transform decomposes the block into visual frequencies — low frequencies represent uniform areas, high frequencies represent edges and fine details.
Then comes quantization: the precision of high frequencies is reduced in a controlled way, while low frequencies are preserved. The quality slider you choose (50, 75, 85%) controls how aggressive this reduction is. At 80–85%, artifacts are invisible in nearly all real-world uses. Below 60%, blocky distortions appear at edges — especially visible on text or logos over flat backgrounds.
JPEG has no transparency support (alpha channel). For transparent backgrounds, PNG or WebP are required.
PNG: lossless, but still compressed
PNG discards no information. Yet a compressed PNG can be far lighter than a raw image of the same dimensions.
The trick: PNG first applies prediction filters to each row of pixels. Rather than storing the raw value of each pixel, it stores the difference between a pixel and its neighbors. These deltas are much smaller than absolute values, and the DEFLATE algorithm (the same as ZIP) compresses them very efficiently.
PNG excels on images with large uniform color areas: logos, icons, screenshots, graphics with text. It performs poorly on photographs — a photo full of subtle color variations generates high deltas that compress poorly.
WebP: designed for the modern web
WebP was developed by Google in 2010. It exists in two modes — lossy and lossless — covering the use cases of both JPEG and PNG in a single format.
In lossy mode, WebP uses neighbor-block prediction rather than DCT. This more sophisticated algorithm produces fewer block artifacts at equivalent file sizes. At identical visual quality, a lossy WebP is typically 25–35% lighter than an equivalent JPEG.
In lossless mode, WebP applies spatial reference compression that exploits repetition within the image — a more elaborate mechanism than PNG's DEFLATE. A lossless WebP is typically 20–30% lighter than an equivalent PNG.
The only significant limitation of WebP: it's not supported by all email clients (notably Outlook) or some legacy editing software.
The right setting for each use case
Photos for the web or social media: WebP at 80–85%, or JPEG at 80% if compatibility is the priority. At this setting, the file is 60–80% lighter than an uncompressed original, with no visible difference on a standard screen.
Logos and graphics with transparency: PNG or lossless WebP. Compressing a logo as JPEG introduces artifacts at sharp edges.
Email attachments: JPEG at 75–80%. WebP is best avoided as an email attachment due to lack of universal support.
Website images: WebP first, with a JPEG fallback for legacy browsers. Core Web Vitals scores reward lighter images — format choice is one of the most direct levers.
How Zipero approaches compression
Zipero's image compressor analyzes your image to identify areas of high visual importance — faces, text, sharp edges — and applies less compression there than to backgrounds and uniform zones. Your image loses weight precisely where it's not noticeable.
Everything runs in your browser via WebAssembly. Your files never leave your device — no upload, no trace.