JPG vs PNG vs WebP: we measured 6 photos and 2 screenshots
Real file sizes and SSIM scores for 6 photos and 2 screenshots saved as JPG, PNG and WebP in the browser, including WebP vs JPEG at equal quality.
Published: 2026-11-12 · Updated: 2026-11-12 · SwiftlyKit
What we tested and how
Most format comparisons repeat the same claims. We wanted numbers from the code our own tools run, so we encoded six photos and two screenshots with the encoder behind WebP Converter and PNG ↔ JPG Converter. Each image was saved at its original size as JPEG and WebP at quality 60, 75, 85, 90 and 92%, and as PNG. In a second pass we first scaled each photo to 1600 px on its long side, a common size for web pages.
Everything ran on 2026-11-12 in headless Chromium 158 on one desktop PC (AMD Ryzen 7 9800X3D, Windows 11). We called the same worker code the site runs in your browser, so the output bytes are the ones the tool would give you in Chromium.
Quality is measured with SSIM, a standard similarity score, computed on brightness (luma) at full resolution. 1.0 means identical to the source; lower means more change. The reference is the source exactly as the encoder received it, after Chromium decoded it and converted its colours to sRGB. Sizes use 1 KB = 1,000 bytes and 1 MB = 1,000,000 bytes.
- Landscape: mountains at sunset with a large, smooth sky, 5184 × 3456 px, 4.15 MB, Adobe RGB colour profile.
- Street: Amsterdam brick facades with fine detail, 4592 × 3448 px, 9.53 MB.
- Portrait: close-up of a person, 4272 × 2848 px, 7.78 MB.
- Night: a Budapest bridge at night, ISO 125 long exposure with little noise, 3580 × 2000 px, 5.28 MB.
- Product: a black smartwatch shot on a phone, 3024 × 4032 px, 2.98 MB, Display P3 colour profile.
- High-ISO: a starry sky at ISO 3200 with visible sensor noise, 4000 × 2670 px, 13.4 MB.
- Screenshots: the SwiftlyKit home page on a 1440 × 900 desktop window, and a tool page on a phone-sized window (1170 × 2532 px).
PNG is the wrong format for photos
Saved as PNG, every photo became 10 to 21 times larger than the JPEG at 85% (9.6× to 20.5× exactly). The PNG files ran from 14.6 MB to 34.2 MB, 1.8 to 5.4 times the size of the camera originals. PNG kept every pixel identical (SSIM 1.0), but a photo does not need that for display.
The table also shows the first trap in format comparisons. At the same quality number, WebP was much smaller than JPEG for five photos, but its SSIM was lower in all five. The two files are not equally good, so the size gap overstates what WebP saves.
| Photo (full size) | Original | PNG | JPEG 85% (SSIM) | WebP 85% (SSIM) |
|---|---|---|---|---|
| Landscape | 4.15 MB | 22.3 MB | 1.18 MB (0.974) | 0.37 MB (0.961) |
| Street | 9.53 MB | 34.2 MB | 3.17 MB (0.984) | 2.43 MB (0.981) |
| Portrait | 7.78 MB | 17.8 MB | 0.87 MB (0.991) | 0.40 MB (0.980) |
| Night | 5.28 MB | 14.6 MB | 1.26 MB (0.976) | 0.99 MB (0.974) |
| Product | 2.98 MB | 15.9 MB | 1.33 MB (0.947) | 0.87 MB (0.938) |
| High-ISO | 13.4 MB | 23.9 MB | 2.48 MB (0.907) | 2.81 MB (0.955) |
WebP vs JPEG at equal quality
Each encoder's quality number means something different, so we also compared the formats at equal SSIM: for each JPEG result, how many bytes WebP needs to reach the same score. We estimated that by interpolating between the two nearest WebP settings we measured.
At 1600 px, WebP was smaller at every point we could compare, by 7 to 31%. The table shows the comparison against JPEG at 75%. At full camera size, WebP was 4 to 28% smaller at all comparable points but one: for the portrait near SSIM 0.985, WebP needed about 15% more bytes. On very clean or smooth images, WebP's score stops rising at higher settings because it smooths fine grain that SSIM counts as detail. For several full-size points on the landscape and portrait, no WebP setting between 60% and 92% reached the JPEG's score at all.
| Photo at 1600 px | JPEG 75% | SSIM | WebP for the same SSIM | Difference |
|---|---|---|---|---|
| Landscape | 27.5 KB | 0.984 | ≈ 25.4 KB | −8% |
| Street | 423 KB | 0.964 | ≈ 302 KB | −29% |
| Portrait | 86.9 KB | 0.981 | ≈ 66.8 KB | −23% |
| Night | 217 KB | 0.969 | ≈ 170 KB | −21% |
| Product | 104 KB | 0.970 | ≈ 81.4 KB | −22% |
| High-ISO | 168 KB | 0.892 | ≈ 119 KB | −29% |
Screenshots: WebP wins clearly, and re-saving as PNG backfires
Screenshots behave differently from photos. At the same quality number WebP was 43 to 53% smaller than JPEG and still scored higher. At equal SSIM it needed 60 to 66% fewer bytes than JPEG at 85, 90 and 92%. At lower JPEG settings the comparison has no estimate, because even WebP at 60% scored higher than those JPEGs.
Re-saving a screenshot as PNG in the browser made it bigger: 1.06 MB instead of the 0.59 MB file we started with, and 0.52 MB instead of 0.30 MB. Chromium's canvas PNG encoder compresses less hard than the program that saved the original. If you need a lossless screenshot, keep the original PNG rather than converting it again.
A high SSIM does not prove that small text is clean. JPEG and WebP can leave faint ringing around letters that this score barely registers, so zoom in on text before you publish.
| Screenshot | Original PNG | PNG re-saved | JPEG 85% (SSIM) | WebP 85% (SSIM) | WebP at JPEG 85% SSIM |
|---|---|---|---|---|---|
| Desktop, 1440 × 900 | 594 KB | 1,057 KB | 124 KB (0.991) | 66 KB (0.996) | ≈ 47 KB (−62%) |
| Phone, 1170 × 2532 | 300 KB | 520 KB | 202 KB (0.996) | 100 KB (0.998) | ≈ 68 KB (−66%) |
Why the same quality number misleads: the high-ISO photo
The noisy night-sky photo reversed the usual result. At 85%, the WebP file was 13% larger than the JPEG (2.81 MB vs 2.48 MB). Read alone, that suggests WebP is worse for noisy photos. But the WebP scored 0.955 against the JPEG's 0.907, so it kept much more of the image. Matched on SSIM, WebP was 24 to 28% smaller at full size and 27 to 31% smaller at 1600 px.
The lesson applies to every format comparison: compare files of equal quality, or at least look at both, before trusting a size difference. A slider value is a setting for one encoder, not a common scale.
Encoding time
WebP took longer to encode. At full size, one photo took 0.56 to 1.49 s as WebP against 0.12 to 0.20 s as JPEG. At 1600 px it was 81 to 218 ms for WebP against 27 to 44 ms for JPEG. These are medians of three runs and include starting the worker. They come from a fast desktop processor, so phones and older laptops will be several times slower; use them to compare formats, not to predict your own wait.
What we recommend
Based on these results, with the limits listed below:
- Save photos as JPEG or WebP, not PNG. PNG photos were 10 to 21 times larger than JPEG at 85%.
- Resize before anything else. The street photo at JPEG 75% was 2.38 MB at full size and 423 KB at 1600 px; the landscape went from 627 KB to 27.5 KB. Image Resizer does this in the browser.
- For websites, WebP Converter saves roughly a fifth to a third on photos at equal SSIM in our 1600 px test (7 to 31%), not the 50 to 70% that same-number comparisons suggest.
- For screenshots and interface images, WebP gave the biggest saving. Avoid JPEG when the image has small text, and keep the original PNG if it must stay lossless.
- Keep JPEG when compatibility matters, for example for upload forms or older software that rejects WebP. PNG ↔ JPG Converter creates JPEG files, and Image Compressor keeps a JPEG as JPEG.
- Whatever you choose, check the result at 100% zoom where the image will be used.
Limits of this test
Eight images are a small set, and the results depend heavily on content: smooth skies and clean portraits compress far more than fine detail or noise. Treat the figures as ranges for similar images, not as a universal rule.
- One machine and one browser engine. Firefox and Safari use different encoders, so sizes and scores will differ. In our code path, Safari cannot create WebP files at all.
- SSIM and PSNR are objective proxies, not a viewing study. On flat skies SSIM penalises lost grain that viewers rarely notice.
- Browser re-encoding drops EXIF and ICC metadata and converts wide-gamut photos to sRGB. That affected the landscape (Adobe RGB) and product (Display P3) photos; SSIM here measures compression loss after that conversion, not the colour change.
- Chromium's JPEG encoder stored colour at reduced resolution (4:2:0 chroma subsampling) at every setting below 100%.
Sample files and credits
The six photos are public domain (CC0) on Wikimedia Commons, and each download was checked against the checksum Commons publishes. CC0 does not require attribution, but we credit the photographers anyway; the file pages are listed under Sources.
The two screenshots show SwiftlyKit pages, captured for this test from a local build of the site.
Try the related SwiftlyKit tools
Sources and further reading
Night photo: “Széchenyi Chain Bridge in Budapest at night.jpg” by Wilfredor, CC0, Wikimedia Commons
Product photo: “Withings ScanWatch 2 black.jpg” by Susie Felber, CC0, Wikimedia Commons