The image format dataset
One JSON file describing what 12 image formats can and cannot do, where each of them is specified, and which conversions between them are possible in a browser. Free to use, no key, no rate limit, CC BY 4.0.
There is no conversion API, and there will not be one
Worth saying up front, because it is the first thing a developer looks for. A REST endpoint that converts an image means the image travels to a server, and not doing that is the entire point of ImageConvert.tools. There is no hosted conversion API here and adding one would make the site's central claim false.
What can be published without breaking anything is the knowledge. Working out which format supports an alpha channel, which can animate, which survives an EXIF round trip and which conversions are even possible turns out to be a research task every time, and the answer does not exist anywhere in a form a program can read. So here it is.
The endpoint
GET https://imageconvert.tools/api/formats.json
Served as application/json with Access-Control-Allow-Origin: *, so it works from a browser as well as from a script. Version 1.1.0, last changed 2026-11-06.
const res = await fetch("https://imageconvert.tools/api/formats.json");
const data = await res.json();
// Which formats can carry an alpha channel?
data.formats.filter(f => f.supports.transparency).map(f => f.label);
// Can this converter turn a HEIC into a PDF?
data.formats.find(f => f.id === "heic").converter.convertsTo.includes("pdf");What is in it
Per format:
mimeTypeandextensionsspecification, the name and URL of the document that defines the formatintroduced, the year it was first published or standardisedcompression: lossless, lossy, lossy or lossless, uncompressed, or containersupports: transparency, animation, colour profiles, EXIF, vectorconverter: whether this site can read it, write it, and the full list of targets it converts to with the page URL for eachnotes, one paragraph of the context a boolean cannot hold
At the top level, a browserSupport block holds the measurements in the table below, with the date, the browser builds, the test files and the method.
Plus a pipeline block naming the decoder and encoder used for every format, and a caveats array. Those caveats are in the file rather than only on this page on purpose: a dataset that travels without its caveats gets quoted without them.
The distinction that matters
supports describes the format. converter describes this implementation. They are not the same question, and conflating them is the most common error in format comparison tables. PNG can carry EXIF; this converter does not write any. GIF can animate; this converter writes a single frame. Read how it works for the full account of what the pipeline keeps and what it drops.
What is deliberately not in it
No file sizes, no compression ratios, no "X percent smaller" numbers. Publishing those would require a documented test set, a stated encoder version and a repeatable method, and until that exists any figure here would be decoration. The one percentage in the dataset is Google's own published WebP comparison, and it is labelled as Google's in the notes field rather than presented as a measurement of ours.
Browser support, tested
Which formats a browser can open by itself and which it can save from a canvas, measured rather than copied from a compatibility table. Measured on 2026-11-06 in headless Chromium 151.0.7922.34, Firefox 142.0.1, WebKit 26.0, on Windows.
| Format | Chromium 151.0.7922.34 | Firefox 142.0.1 | WebKit 26.0 |
|---|---|---|---|
| PNG | decode yes encode yes | decode yes encode yes | decode yes encode yes |
| JPG | decode yes encode yes | decode yes encode yes | decode yes encode yes |
| JFIF | decode yes encode n/a | decode yes encode n/a | decode yes encode n/a |
| WebP | decode yes encode yes | decode yes encode yes | decode yes encode yes |
| AVIF | decode yes encode no | decode yes encode no | decode no encode no |
| GIF | decode yes encode no | decode yes encode no | decode yes encode no |
| BMP | decode yes encode no | decode yes encode yes | decode yes encode no |
| TIFF | decode no encode no | decode no encode no | decode no encode no |
| ICO | decode yes encode no | decode yes encode no | decode yes encode no |
| SVG | decode <img> only encode n/a | decode <img> only encode n/a | decode <img> only encode n/a |
| HEIC | decode no encode no | decode no encode no | decode no encode no |
Decode means a small test file loaded as a Blob, into an <img> element and into createImageBitmap; "<img> only" means the second failed. Encode means canvas.toBlob returned a file of exactly the requested type, because a browser without the encoder quietly hands back a PNG instead. JFIF shares the JPG encoder and SVG is vector, so neither has an encode result.
The WebKit column is Playwright's WebKit build on Windows, not Safari: Safari on macOS and iOS can also use system codecs, HEIC among them. Where a browser cannot do something, this converter brings its own code where it can: HEIC and TIFF are read, and AVIF, GIF, BMP, ICO and PDF are written, without the browser's help. The pipeline block in the file names each decoder and encoder.
Licence
Creative Commons Attribution 4.0. Use it commercially, modify it, redistribute it. The one condition is attribution:
Image format capability dataset, ImageConvert.tools, https://imageconvert.tools/api
Corrections
Every entry was written from the specification named in it. If you find something wrong, tell us and it gets fixed, with the version number bumped. A dataset nobody can correct is not worth citing.
Coverage
11 formats can be read: PNG, JPG, JFIF, WebP, AVIF, GIF, BMP, TIFF, ICO, SVG, HEIC. 9 can be written: PNG, JPG, WebP, AVIF, GIF, BMP, TIFF, ICO, PDF. The three that are read but not written, and the reasons why, are set out on the conversion hub.