Compare image formats
Compare image formats on your own picture, free and in your browser with no upload. The image is encoded as WebP, JPEG, PNG, GIF and more, with the file size, encode time and a size-against-quality curve for each.
General advice about formats is guesswork about your image. This measures yours, in this browser, and shows the size against the quality for each.
In detail
Compare formats encodes one image in every format this browser can write, including WebP, PNG, JPEG, GIF, BMP, QOI, TGA and PPM, and reports the file size, the encode time and, for the lossy ones, a curve of size against quality. JPEG and WebP are each measured at five quality points: 0.5, 0.7, 0.8, 0.9 and 1.0.
The encoding happens in your own tab, so the answer is about your image on your machine rather than general advice. AVIF is listed as not measured, because no browser's built-in encoder can write it and File4Edit has no AVIF encoder of its own. In Safari, WebP is measured with File4Edit's own encoder, so the encode time there is that encoder's. Select any row to see it beside a lossless render and download the exact file measured.
Compare image formats at a glance
- Input
- JPG, PNG, WebP, AVIF, HEIC, GIF (first frame), BMP, TIFF, SVG and other image formats
- Output
- A comparison table, and any measured file as a download
- What it does
- Encodes WebP, PNG, JPEG, GIF, BMP, QOI, TGA and PPM
- Size and encode time per format
- A quality curve for JPEG and WebP
- A reduced test for quick results
- A split view against a lossless render
- Download the measured file
- Who it is for
- Web developers and designers deciding which format to ship.
- Limits
- AVIF cannot be measured, because no browser's built-in encoder can write it and File4Edit has none of its own.
- In Safari and on iPhone and iPad, WebP is measured with File4Edit's own encoder, so its encode time there is not the browser's.
- Encode times vary between runs and machines.
- Privacy
- Runs entirely in your browser: the file is never uploaded, and nothing is kept after you close the tab.
How to compare image formats
- Drop an image Measuring starts as soon as the image decodes. The formats are encoded one at a time on purpose — concurrent encodes would compete for the same cores and give every timing a worse number than it deserves.
- Leave the reduced test on for the first pass It encodes a 1200px render instead of the full image, which turns minutes into seconds. The formats keep their order and roughly their ratios at full size; the byte counts do not. Turn it off when you want the bytes you would actually ship.
- Read the table Each row gives the size, the share of the original (or of PNG, when the test was reduced), the encode time and the curve. Rows sort smallest first once the run finishes, and the smallest is marked.
- Move the reading point along the curve Read the curve at 0.5, 0.7, 0.8, 0.9 or 1.0. Every point was already measured, so moving it re-reads the run and re-encodes nothing. Which format wins can change with the quality, and that is the point.
- Select a row and download it Selecting a row shows it in a split view against a lossless render of the same pixels, so the size is set beside what it cost. Download saves the exact file that was measured.
Compare image formats: common questions
Why is AVIF missing from my table?
No browser's built-in encoder can write AVIF yet, whether in Chrome, Edge, Firefox or Safari, and File4Edit has no AVIF encoder of its own and sends your image nowhere to make one, so AVIF is listed under "not measured" with that reason instead of being quietly dropped.
Why are ICO, ICNS, PGM and PBM skipped?
Because their sizes would answer a different question. ICO and ICNS are containers holding a set of icons rather than an encoding of your picture, and PGM and PBM discard color, so all four would come out small for a reason that is not compression. Each is shown in a "not measured" list with its reason.
How exact are the encode times?
They are wall-clock times taken in this tab and they include preparing the pixels for the codec, so they move between runs and between machines. Use them to rank the formats against each other within one run, not as an absolute benchmark.