Benchmarks
What our tools actually did to real files, measured rather than promised — including the conversions that make files bigger.
How these numbers are produced
Every job this site runs records the input size, the output size and how long it took. The table below aggregates those records. Nothing here comes from a curated sample set chosen to look good — it is whatever people actually uploaded.
Three rules keep the figures honest, and all three cost us something:
- Median, not average. One unusual file drags an average a long way. The median describes the typical case; an average can describe a case that never happened.
- Nothing published below 10 samples. A tool that has run twice has two data points, not a statistic. 15 tools are being measured but not yet shown for this reason.
- Results that go the wrong way are printed anyway. Converting from a lossy format to a lossless one — WEBP to PNG, AVIF to JPG — makes files several times larger. That is what the format change means, and you would find out regardless; better here than after the download.
Sample counts only ever grow. Uploads are deleted within two hours and results within twenty-four, but the measurement itself — sizes and timing, with no file name and nothing identifying who sent it — is kept so this page can exist.
Measured results
399 conversions measured in total. Median size change is relative to the file you started with.
| Tool | Median size change | Median time | Typical input | Samples |
|---|---|---|---|---|
| PNG to JPG | 89% smaller | 2.6 s | 2.1 MB | 191 |
| Compress PNG | 84% smaller | 661 ms | 2.8 MB | 56 |
| JPG to WEBP | 4% smaller | 3.9 s | 170 KB | 27 |
| Resize Image | 98% smaller | 349 ms | 1.9 MB | 22 |
| PNG to WEBP | 98% smaller | 1.7 s | 2.8 MB | 21 |
| Compress Video | 81% smaller | 5.9 s | 3.9 MB | 14 |
| Compress JPG | 30% smaller | 1.1 s | 170 KB | 11 |
How to read the size column
A large saving is not automatically a good result and a small one is not a failure. What the number tells you is how much slack was in the original file:
- Big savings usually mean the source was stored inefficiently — a screenshot kept as PNG, or a photo exported at a quality far above what anyone can see.
- Small savings mean the file was already well compressed. Running it through again trades quality for almost nothing, which is why our compression tools return your original untouched when the result comes out larger.
- Growth is expected whenever the target format is lossless. You are buying an exact, re-editable copy, and it costs bytes.
The quality side is measured too. Rather than applying a fixed quality number, our image tools encode, compare the result against your original with a structural similarity measurement, and re-encode until it clears the threshold for the mode you picked — 0.998 for Best, 0.99 for Balanced, 0.98 for Smallest. Each tool page explains what that means for that specific conversion under Output quality.