The only honest way to choose a format is to try them all
Every guide to image formats gives you the same table: PNG for graphics, JPEG for photographs, WebP for the web. The advice is broadly right and it will still mislead you regularly, because the numbers depend on the picture in front of you and nothing else.
A screenshot of text is enormous as a JPEG and tiny as a PNG. A photograph is the reverse. A logo with one soft shadow behaves like neither. Two pictures the same size on screen can differ by a factor of ten in the same format.
So this page does not guess. It encodes your actual picture in every format your browser can write, and shows you the real byte counts together, before you choose. Here is what that looked like for a 600×400 graphic with a transparent hole, measured while this tool was built:
The smallest file is not automatically the right answer. The JPEG above is half the size of the PNG and it destroyed the transparency to get there. That is why the comparison marks which formats keep see-through areas and which cannot, and why it recommends the smallest one that keeps what your picture actually has, rather than the smallest one outright.
What each format is genuinely good at
PNG stores every pixel exactly, as the specification requires, and carries a full transparency channel. That makes it right for screenshots, logos, diagrams, line art and anything with flat colour or hard edges — and expensive for photographs, where storing every subtle gradation faithfully costs far more than your eye can use.
JPEG was designed for photographs in 1992 and is still very good at them. It gets small by discarding detail the eye barely registers, which is exactly the wrong trick for sharp edges — text saved as JPEG grows fuzzy halos. It has no transparency channel at all, a limitation of the original standard that has never changed.
WebP does both jobs. It has a lossy mode for photographs and a lossless mode for graphics, keeps transparency in both, and Google’s own figures put it around 25–34% below comparable JPEG. Every current browser reads it, as caniuse records. For most web work it is simply the better choice, and it is usually what the comparison here recommends.
AVIF is newer, built on the AV1 video codec, and typically smaller again — sometimes dramatically so on photographs. Support for reading it is now good; support for writing it from a browser is patchier and slower. If yours cannot, AVIF simply does not appear in the list rather than being offered and quietly failing.
What happens to transparency, and why it is usually a nasty surprise
This is the single most common complaint about converting images, and it is almost always the same story: a logo with a see-through background is converted to JPEG, and the background comes back solid white. Sometimes solid black, which is worse and looks like a bug.
It is not a bug. JPEG has nowhere to put transparency, so those pixels must become some colour. The black version happens when a tool draws the picture onto a blank canvas and encodes without filling it first — unset canvas pixels are transparent black, and flattening transparent black gives you black.
This page handles it three ways. It detects transparency in your file and says so. It warns before converting, naming the format and the colour that will replace it. And it lets you choose that colour — white, black, or anything you pick — so a logo destined for a white page and one destined for a dark header can both come out right.
If you would rather not lose it at all, the comparison makes that easy: WebP is usually both smaller than the JPEG and able to keep the transparency, which is a rare case of the better answer also being the cheaper one.
What a conversion always costs you, whatever the format
Two things go every time, and no browser-based converter can avoid either.
The metadata. Converting redraws the picture, and what comes back is pixels alone. The camera model, the lens, the date, the copyright field and often the exact GPS coordinates are all gone. For sharing a photograph publicly that is a genuine privacy benefit; for a photographer cataloguing a shoot it is a loss. Either way it should be stated rather than discovered.
Anything the previous encoder already threw away. Converting a JPEG to PNG does not restore detail — it faithfully preserves the JPEG’s compression artefacts in a much larger file. This is the most common wasted conversion there is. PNG cannot undo lossy compression; nothing can.
And one thing that goes only sometimes: animation. An animated GIF converted through a browser canvas keeps only its first frame, because a canvas holds one frame at a time. The row says so rather than handing you a still picture and letting you find out.
Questions people ask about converting image formats
Which image format is smallest?
For most pictures, AVIF then WebP then JPEG then PNG — but “most” is doing a lot of work in that sentence. A screenshot of text inverts the order completely. That is precisely why this page measures your file instead of quoting the table.
Is converting PNG to JPG going to lose quality?
Yes, a little, and irreversibly — JPEG is lossy by design. On a photograph you will not see it at a sensible quality setting. On text, screenshots or line art you will, as soft halos around every sharp edge. If the picture has hard edges, prefer PNG or lossless WebP.
Can I convert back afterwards?
You can convert the file, but you cannot recover what was discarded. JPEG to PNG gives you a large, faithful copy of the already-degraded JPEG. Always convert from the best original you have, not from a converted copy.
Will my transparent background survive?
With PNG, WebP or AVIF, yes. With JPEG, no — it becomes the colour you choose above, and the tool warns you first. Every preview here sits on a chequerboard so you can see the transparency rather than take it on trust; on a white background, a see-through area and a white area look identical.
Why is AVIF missing from my list?
Because your browser cannot write it. Reading AVIF is widely supported; encoding one from a web page is not, and it is slow where it works. The tool tests the browser at startup and only offers what actually comes back correctly — which also protects you from a subtler failure, where a browser accepts a request for one format and silently returns another.
Can I convert several files at once?
Yes, up to 30, all to the same format. The comparison is measured on the first one to keep things quick; the rest are converted the same way, and each row shows its own before-and-after size.
Does this work with HEIC photos from an iPhone?
Only if your browser can decode HEIC, which most desktop browsers cannot. If a file will not load, that is why. Safari on Apple hardware generally manages it; elsewhere you will need the HEIC to JPG Converter.
What quality setting should I use?
Somewhere between 75 and 85 for photographs is the usual sweet spot — below about 70 the artefacts start to show on smooth gradients like skies. The slider here goes down to 10 for when you genuinely need a tiny file and can accept how it looks. Lossless formats ignore it entirely, which is why it disappears when you choose PNG.
Are my images uploaded?
No. They are read from disk into this page, encoded here, and handed straight back. Nothing is sent anywhere and nothing is stored.
Where the facts on this page come from
- MDN — image file types and formats, on which formats are lossy, which are lossless and which carry transparency.
- PNG Specification (Third Edition), W3C, and ITU-T T.81, the original JPEG standard.
- WebP, Google, for the lossy and lossless modes and the 25–34% comparison against JPEG.
- AV1 Image File Format (AVIF), Alliance for Open Media.
- caniuse — WebP and caniuse — AVIF, for browser support.
- MDN — HTMLCanvasElement.toBlob(), which is what actually does the encoding here, and which falls back to PNG when asked for a type the browser cannot write.
