JPG to PNG Converter — Convert JPEG to PNG Online

What this converter does, and the 1 myth it corrects

This tool re-encodes a JPG image as a PNG entirely inside your browser: the file is decoded to pixels, drawn onto an HTML canvas at its own width and height, and saved through the browser's built-in Portable Network Graphics encoder. Nothing is uploaded; no server sits between you and it. The most important thing to say up front is what it does not do. It does not add transparency. A JPEG file has no alpha channel at all, so each pixel it stores is solid, and re-saving the exact pixels leaves all 4 RGBA channels opaque, alpha pinned at 255. We verified this directly: after conversion, every 1 of the output pixels reads alpha 255, the value for solid, 255 of a possible 255, not the 0 that means see-through. If you have read that this kind of converter makes the background transparent, that claim is wrong at the format level. Transparency has to be painted in afterwards with a dedicated eraser or an editor, which sets chosen pixels to alpha 0; this converter only changes the container.

Lossy JPEG versus lossless PNG

JPEG and PNG compress in opposite ways. The Joint Photographic Experts Group format, JPEG, from 1992, uses lossy compression built on the Discrete Cosine Transform, or DCT: it splits each frame into blocks of 8 by 8 pixels, converts each to 64 frequency coefficients via an 8 by 8 transform, and quantises away the ones the eye barely notices, often discarding 80–90% of them. That is why JPEG stays small and why text and hard edges pick up faint block artifacts. Portable Network Graphics, from 1996, uses lossless DEFLATE, the same LZ77 plus Huffman coding as a ZIP file from 1993, storing all 3 colour channels of every pixel exactly, 8 bits each for 24-bit colour. Nothing is discarded, so lines and solid colours stay crisp. The trade is size: because PNG keeps everything, a photograph saved this way is usually multiples larger, while busy images hit 10 to 20 times the JPEG, occasionally 30.

Why the size jumps, and why you see it

Converting almost always yields a bigger output, by a wide margin. In 1 live test a 6 kilobyte JPG became a 130 kilobyte PNG, a 21x jump, because DEFLATE cannot compress photographic noise the way the DCT can. That is expected, not a bug, which is exactly why the tool prints the original size, the new size, then the multiplier for every file, plus a batch total. You see the real cost before you download, instead of discovering a 20 megabyte folder later. On a flat graphic with few colours the jump is far smaller, and the format is genuinely the right choice; with a photograph, the honest question is whether you need lossless at all, given the 5 to 20x size cost.

When the conversion is the right move

There are real reasons to make this conversion, and they are about editing, not shrinking. The first is stopping generation loss, the quality decay of repeated lossy saves. Every time you open a JPG, edit it, and re-save as JPG, the lossy pass runs again so quality drops a little; after 5 to 10 rounds the degradation is clearly visible, each pass shedding another 5–10% of detail. Save as PNG once and every later edit is lossless, so you can crop, annotate, then re-export as many times as you like with 0 added loss. The second is preparing an image for transparency work: a background remover needs a format that can store an alpha channel, and PNG can while JPG cannot, so converting first gives the eraser somewhere to write the see-through pixels. The third is sharp-edge content, screenshots of text, diagrams, and UI mockups, where JPEG's block artifacts look ugly while PNG stays pixel-crisp.

How PNG stores an image, in 5 chunk kinds

A PNG file is a stack of typed chunks, and understanding them explains its size. After an 8-byte signature come 4 critical chunk types: IHDR holds the header, width, height, bit depth, and 1 of 5 colour types in exactly 13 bytes; PLTE holds an optional palette of up to 256 colours; IDAT holds the DEFLATE-compressed pixel data, usually 90%+ of the file; and IEND marks the end in 0 data bytes, a fixed 12-byte chunk, 4 length plus 4 type plus 0 data plus 4 CRC. A 24-bit truecolour image like a converted photo skips the palette and stores full 24-bit or 32-bit pixels in IDAT, which explains its size. An 8-bit palette PNG, by contrast, caps at 256 colours in 1 byte per pixel and can be tiny, so logos and flat graphics compress far better than photographs in the exact same format.

The conversion pipeline, step by step

Under the hood the process is short. The browser decodes the JPEG binary, undoing its DCT to recover a full grid of red, green, blue values at 8 bits each. That grid is painted onto an offscreen canvas sized to the image. The tool then calls the canvas toBlob method with the image or png type, then its encoder runs DEFLATE over the raw pixels and wraps them into the chunk structure, an 8-byte signature followed by the IHDR, IDAT, and IEND chunks, 3 of 4 mandatory ones. Because the source carried no transparency, the encoder writes a fully opaque image; the pixels are preserved bit for bit from the decoded JPEG, so no quality is lost in this direction. The whole pass takes under 1 second on a typical 2 to 5 megapixel image, and never leaves your machine, even for a 4000 by 3000 photo of 12 megapixels.

JPEG versus PNG by the numbers

A few fixed figures explain most behaviour. JPEG was published in 1992 and uses 8 by 8 blocks, 64 coefficients per block, and a quality scale from 1 to 100, where 75 to 85 is typical and 90+ is near-lossless. PNG was published in 1996, became ISO standard 15948 in 2004, and offers 5 colour modes and 5 bit depths of 1, 2, 4, 8, and 16 bits. Its DEFLATE stage runs at 1 of 10 compression levels, 0 through 9. A 24-bit truecolour PNG stores 3 bytes per pixel, rising to 4 bytes or 32 bits with alpha, so a 1920 by 1080 frame holds about 2.07 million dots and roughly 6.2 megabytes uncompressed, 2.07 million pixels times 3 bytes, before DEFLATE. That same frame at quality-80 JPEG might be 300 kilobytes, making the lossless version 3 to 10 times heavier. With a small graphic under 16 colours the numbers invert: an 8-bit palette PNG can undercut it, because lossy coding wastes bytes fighting the sharp edges a 256-colour palette, just 768 bytes, stores for free.

Quick reference

• Direction: JPG or JPEG into PNG, re-encoding in the browser canvas.
• Transparency: not added; a JPEG has no alpha, so the PNG comes out fully opaque at alpha 255.
• Compression: PNG is lossless DEFLATE; the source JPEG was lossy DCT.
• File size: PNG is usually larger, often several times, and up to 10 or 20 times for a photo.
• Size readout: original size, PNG size, and multiplier are shown per file, plus a batch total.
• Best use: stopping generation loss, prepping for transparency work, or sharp-edged graphics.
• Quality: the decoded JPEG pixels are preserved exactly; this direction adds 0 further loss.
• Privacy: 100 percent in-browser; no upload, no server.