PNG • lossless
Often useful for icons, screenshots and detailed graphics. Usually a larger file for complex photos, but not always.
Is WebP always smaller? Is PNG better for words and sharp lines? The answer depends on the picture and the browser. Run the three test scenes below, or select an image from your device. This tool creates actual output files at matching dimensions, then measures their real byte sizes.
Unlike a static format comparison chart, the numbers here come from your browser's encoder. You can download every result and inspect the differences yourself.
Run a comparisonEach output uses exactly the same rendered pixels and dimensions. The PNG baseline is lossless; JPEG and WebP use the selected browser quality setting. Browser and platform differences are expected.
Often useful for icons, screenshots and detailed graphics. Usually a larger file for complex photos, but not always.
Widely compatible for photographic material. Fine text or hard boundaries can develop compression artifacts.
Modern format supported by many browsers. Savings are not guaranteed and depend on scene detail and encoder.
Dimensions and output details will appear after you run a test.
We ran the three built-in scenes in headless Firefox 153.6 ESR on Linux on November 11, 2026. The figures below are the actual encoded file sizes from that test, not advertised conversion rates. Every image in a row used the same canvas pixels and dimensions. Browser updates, hardware and encoders can produce different numbers.
| Test scene | JPEG/WebP quality | PNG | JPEG | WebP |
|---|---|---|---|---|
| Text and sharp edges | 80% | 119.7 KB | 36.9 KB | 12.3 KB |
| Smooth gradients | 55% | 370.7 KB | 22.3 KB | 6.3 KB |
| Fine color texture | 90% | 199.3 KB | 694.9 KB | 299.7 KB |
What this means: WebP produced the smallest file for the text and gradient samples in this Firefox test. But PNG was smaller for dense synthetic texture at 90% quality. Smaller is not necessarily visually better. To decide what to publish, run the sample yourself and compare sharpness, banding, compatibility and file size on your own device. These results do not predict the output of every real photograph.
A test scene is drawn to a Canvas with the same dimensions for each format. Uploaded images are decoded locally, reduced only when needed, and centered on a solid white background. That prevents a misleading comparison where one format includes transparency and JPEG cannot.
The page uses the browser's canvas.toBlob() encoder, reads each resulting file's byte length and compares it with that test's PNG. The percentage shown means smaller or larger than the PNG result, not smaller than the original upload.
Text and geometric edges reveal ringing or blur. Gradients can show banding. Fine texture challenges the compressor with high-frequency detail. Try each one; a format that wins one scene may lose another.
This is a visual comparison and size measurement, not an objective quality score. It does not measure PSNR or SSIM, preserve EXIF metadata, animated frames, ICC profiles or transparency. Safari, Firefox and Chrome may encode different file sizes at the same percentage.
Different encoders make different trade-offs, especially at lower quality settings or on simple drawings. There is no universal size winner. Choose the result that remains visually acceptable for your specific image.
No. A smaller byte count can come from visible loss of detail or color accuracy. Inspect at full size and on a phone before publishing. Save the original file separately.
For a like-for-like test, this lab places images on an opaque white background. It cannot demonstrate transparent-image behavior. Use a transparency-aware workflow when you actually need an alpha channel; JPEG itself cannot represent transparency.
No image content is uploaded by this lab: the drawing, encoding, previews and file downloads run through your browser's own Canvas and Blob APIs. The page itself is served by ALINOVA, but image processing is local.