ReadoutKit

Image converter

Convert between PNG, JPEG and WebP in your browser, with control over what happens to transparency.

Runs entirely in your browser. Nothing you enter is uploaded.

Written and maintained by ReadoutKit / Takashi Yamazaki

How this works

What each format is for

The three formats here are not competitors so much as tools for different jobs, and picking by habit rather than by content is the most common mistake.

FormatCompressionTransparencyBest for
PNGLosslessYesScreenshots, logos, diagrams, anything with sharp edges
JPEGLossyNoPhotographs, maximum compatibility
WebPBothYesAlmost everything on the modern web

The rule that catches most cases: if the image has large areas of flat colour and hard edges, use a lossless format. If it is a photograph with continuous gradients, use a lossy one. A screenshot saved as JPEG develops visible smearing around text; a photograph saved as PNG can be five times larger for no visible benefit.

The transparency problem

Converting to JPEG is the one direction that loses information beyond compression. JPEG has no alpha channel, so the question is not whether transparent pixels change but what they change to.

Many converters answer that silently, and the default in a bare canvas is black — which is why a logo with a transparent background so often arrives as a logo on a black rectangle. This tool asks instead, and defaults to white because that is what most documents and web pages sit on. If your logo will be placed on a coloured background, set that colour here and the edges will blend rather than showing a white halo.

The same picture, in each format

The table above says what the formats are for; this one says what they cost. It is one 1280 × 992 screenshot — flat colour, text edges, no photographic gradients — decoded and written out through the same canvas path this tool uses.

FormatWritten asSizevs original
PNGoriginal file89 KB100%
JPEGlossy, quality 0.8585 KB95.6%
WebPlossy, quality 0.8550.2 KB56.5%
PNGlossless132.4 KB148.7%
WebPlossless281 KB315.8%

Three results are worth pulling out, because each contradicts something people expect of a converter.

Converting to JPEG barely shrank it — 95.6% of the original at quality 0.85, and at 0.95 it goes to 144.6%, meaning the file is half again as large and visibly worse around the text. JPEG is not a compressor you apply to any image; it is a format for photographs, and this image is not one.

WebP won by a wide margin at 56.5%, on an image that is not what lossy formats are supposed to be good at. That is the general pattern: when in doubt, WebP.

Both lossless conversions made it bigger — PNG through the canvas at 148.7%, WebP lossless at 315.8%. The original had been through a PNG optimiser, and a browser's encoder is not one. Converting a well-optimised PNG to another lossless format is not a saving; it is a rewrite at a worse ratio.

Conversions worth avoiding

  • JPEG to PNG, expecting improvement. Lossless storage of an already-degraded image preserves the damage perfectly and multiplies the file size. It does not restore anything.
  • An optimised PNG to lossless WebP. Measured above at more than three times the size. WebP's advantage is on its lossy side.
  • Repeated lossy round trips. Every JPEG or WebP save discards detail. Converting back and forth a few times produces visible artefacts that nothing can undo. Always start from the highest quality original you have.
  • PNG for photographs. Common when a screenshot tool is used to capture a photo. The result is enormous and looks no better.

The through-line is that a converter changes the container, not the contents. It cannot add detail that a previous save removed, and it cannot make a format suit an image it was not designed for. Choosing by what the picture is made of — flat colour and edges, or continuous gradient — decides almost everything else.

Why AVIF is not on the list

AVIF is the obvious omission, and the reason is worth stating rather than hiding. Browsers candisplay AVIF, but the encoder is a separate matter: asking a canvas for one returns a PNG instead, silently, with no error. A converter that listed AVIF would hand you a file named.avif containing PNG bytes.

That is a worse outcome than not offering it, so the format list here is exactly the set the browser can genuinely produce. If you need AVIF, it has to come from a tool that ships its own encoder — which, on the web, means uploading the image to a server. WebP is the closest thing available without that trade, and it is usually within a few percent of AVIF on photographs.

How the conversion happens here

The image is decoded by the browser, drawn onto a canvas, and re-encoded in the target format — three steps that browsers have supported natively for years. Nothing is transmitted, which means no upload wait, no queue, no file size cap set by someone else's bandwidth bill, and no copy of your image sitting on a server you have never heard of.

One side effect worth knowing: redrawing through a canvas keeps only the pixels. Camera metadata, timestamps, colour profiles and GPS coordinates are not carried across. That is often exactly what you want before publishing an image, but keep the original if any of it matters.

Common questions

Which format gives the smallest file?
WebP, in most cases, and by more than people expect. Measured on a 1280 × 992 screenshot that was 91.4 KB as PNG, writing it out at quality 0.85 gives 51.5 KB as WebP against 87.1 KB as JPEG — the WebP is 41% smaller than the JPEG for the same setting. The exception is a PNG that has already been through an optimiser: converting that to lossless WebP came out at 315.8% of the original.
Will converting to PNG improve a blurry JPEG?
No. PNG stores what it is given without loss, so it preserves the existing damage exactly and makes the file larger doing it. Detail discarded by an earlier lossy save is gone from the file and cannot be recovered by any conversion. Converting is worth doing when you need transparency or a specific format for a tool downstream — not as a repair.
Does converting remove EXIF and GPS data?
Yes. The image is decoded, drawn onto a canvas and re-encoded, and that path carries only pixels. Camera model, timestamps, colour profile and GPS coordinates are not written to the output. That is usually what you want before publishing a photo, but it is irreversible for the converted copy, so keep the original if the metadata matters.
Why did my transparent background turn white?
Because JPEG has no way to store transparency. Every pixel in a JPEG is opaque, so the transparent areas have to become some colour. This tool lets you pick which one and defaults to white; without that step they would come out black, which is the usual unpleasant surprise. If you need the transparency, convert to PNG or WebP instead.
Can it convert HEIC files from my iPhone?
Only if your browser can decode HEIC, which Safari on Apple devices can and most others cannot. Where the browser can read the file, conversion works normally. Where it cannot, the file will fail to load — there is no way for a page to decode a format the browser does not support without shipping a large decoder library.
Does converting lose quality?
It depends on the direction. PNG to WebP or JPEG is lossy, so some detail is discarded. JPEG to PNG is lossless but does not recover anything already lost — you get an exact copy of the degraded image in a larger file. WebP to JPEG re-encodes lossy data that was already lossy, which compounds the loss slightly.
Which formats can my browser write?
The tool checks at load and only offers what actually works. Browsers silently fall back to PNG when asked for a format they cannot encode, which would otherwise produce a file with the wrong extension, so the check happens before you choose rather than after.
Is there a file size limit?
Only what your device can hold in memory. Because there is no upload, there is no server limit and no queue. Very large images — well beyond 50 megapixels — may be slow or may fail on a phone with little free memory.

Sources

The formulas and rates used by this tool come from the documents below. Rules change; when a figure here disagrees with the source, the source is right.

Figures checked against the sources below on September 1, 2026.

Other tools