Skip to content
ALINOVA
☰ Menu
ALINOVA / ORIGINAL BROWSER EXPERIMENT

Image Format Lab: compare PNG, JPEG and WebP using the same pixels.

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 comparison
01 / EXPERIMENT

Change the source and measure for yourself

Each 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.

Each test is generated locally.
Optional; selecting a file takes priority over the test scene.
PNG does not use this quality setting.
Ready. Press Run comparison to generate the three files.
Run a comparison to display the PNG preview.

PNG • lossless

—
Reference sizeDownload PNG

Often useful for icons, screenshots and detailed graphics. Usually a larger file for complex photos, but not always.

Run a comparison to display the JPEG preview.

JPEG • lossy

—
Awaiting a comparisonDownload JPEG

Widely compatible for photographic material. Fine text or hard boundaries can develop compression artifacts.

Run a comparison to display the WebP preview.

WebP • lossy in this test

—
Awaiting a comparisonDownload WebP

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.

02 / ALINOVA RECORDED TEST

A documented comparison: one format does not always win

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.

Measured output sizes from the ALINOVA Image Format Lab in Firefox on November 11, 2026
Test sceneJPEG/WebP qualityPNGJPEGWebP
Text and sharp edges80%119.7 KB36.9 KB12.3 KB
Smooth gradients55%370.7 KB22.3 KB6.3 KB
Fine color texture90%199.3 KB694.9 KB299.7 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.

03 / REPRODUCIBLE METHOD

What the test measures — and what it doesn't

One rasterized source for all formats

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.

Actual output bytes, not estimates

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.

Three source scenes reveal different trade-offs

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.

Limits of the experiment

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.

Repeat the test

  1. Select Text, shapes & sharp edges and run at 80% quality. Compare small lettering and thin borders.
  2. Switch to Smooth color gradients and observe any banding in the darker areas.
  3. Run Fine-grained color texture at both 55% and 90%. Compare resulting sizes and visible artifacts.
  4. Optionally select a photo of your own. Review the generated dimensions before comparing file sizes; large originals may be resized for this demonstration.
  5. Download all three and open them at 100% size in an image editor to inspect details.
04 / COMMON QUESTIONS

Which format should I choose?

Why might JPEG appear smaller than WebP?

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.

Does a smaller file mean an image is better?

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.

What happens to a transparent logo?

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.

Will files be sent to a server?

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.