A PNG made from a JPG keeps every flaw the JPG already had
JPEG throws detail away at the moment a file is saved. The encoder cuts the picture into 8 by 8 pixel blocks, rounds off the fine variation inside each block, and stores what is left. PNG works the other way round. Every pixel is stored exactly, packed with lossless DEFLATE compression.
So a JPG to PNG conversion records the JPG faithfully, block edges and color smearing included. No detail comes back, and no converter on any site restores lost detail. What you gain is a file where every later save is lossless. Re-save a JPG ten times while you edit and each save shaves off a little more. Re-save a PNG ten times and the last file matches the first, pixel for pixel, apart from your edits.
Convert, or leave the JPG alone
Convert to PNG when
- You plan several rounds of cropping, annotating or retouching before the final export.
- An upload form refuses anything but PNG. App store listings, some print shops and plenty of CMS logo fields work this way.
- You want a transparent background and the subject sits on flat white or one other solid color.
- The JPG is a screenshot or diagram you will drop into slides as a working file.
Keep the JPG when
- The image is a photo headed for a web page. The PNG version weighs several times more and looks the same.
- You are emailing camera photos. Attachment limits fill up after two or three PNGs.
- You hope for a sharper picture. A format change fixes no blur, no noise and no block artifacts.
- You are archiving a phone library. Storage needs grow by roughly the same multiple as each file.
Why the PNG comes out heavier, worked through
Take a 4000 by 3000 photo from a phone. Decoded, the picture holds 12 million pixels at 3 bytes of color each, which is 36 MB of raw data. The camera JPEG squeezes those pixels into roughly 3 to 5 MB by discarding grain the eye barely notices. DEFLATE has no permission to discard anything, and photographic grain looks like random noise to a lossless compressor, so the PNG usually lands between 15 and 25 MB.
| Source JPG | Typical JPG size | Typical PNG size | Growth |
|---|---|---|---|
| Phone photo, 12 MP | 3 to 5 MB | 15 to 25 MB | 4x to 6x |
| Product shot on white, 2000 px | 200 to 400 KB | 0.8 to 2 MB | 3x to 5x |
| Screenshot saved as JPG, 1920 by 1080 | 250 to 450 KB | 0.4 to 1.2 MB | 1.5x to 3x |
| Same product shot, background cleared | 200 to 400 KB | 0.5 to 1.4 MB | 2x to 4x |
Those ranges are typical, not promises. The Growth column in the batch rail shows the real multiple for your own files. The setting with the biggest effect is Longest side. Halving width and height leaves a quarter of the pixels, and the PNG shrinks roughly in step. A 12 MP photo scaled to 1920 pixels usually ends up near 3 to 4 MB.
About the quality slider other converters show
PNG has no quality setting. The format is lossless, so there is no detail to trade for bytes. Browser based converters encode with canvas.toBlob(callback, 'image/png', quality), and every major browser ignores the quality argument for PNG. A slider on a JPG to PNG page changes nothing in the output file, which is why this one has none. The two levers with a real effect on PNG size are pixel dimensions, offered above, and color count. For the second, run the finished PNG through a palette quantizer such as pngquant on your desktop.
Clearing a background, the only way a JPG gains transparency
A JPG has no alpha channel. Convert one without touching the background and every pixel in the PNG is fully opaque. Transparency has to come from a decision about which pixels become see-through. Clear a color decides by color distance from one picked pixel.
- Pick the color. Click the background on the preview. The page averages a small patch under your cursor, which smooths out JPEG noise, and switches Background to Clear a color.
- Raise Tolerance. Watch the checkerboard appear behind the subject. Stop when the last specks of background go. Around 20 suits most studio white.
- Soften the edge. If the outline looks jagged, raise Edge softness. Pixels near the cutoff fade out partway instead of vanishing.
- Check every file. The settings apply to the whole batch. Click through the rail before you download, since one photo with a cream backdrop needs a different pick from the rest.
Where color keying fails. A white mug on a white table loses its rim, because the rim matches the key color. Soft shadows and gradient backdrops survive, because they drift away from the key. JPEG also blends the backdrop into every edge pixel, so a product keyed off white keeps a faint pale fringe. On a white page nobody notices. On a dark page the fringe shows. For busy or shaded backgrounds, use the background remover, which separates subject from scene by shape instead of color.
What happens to metadata, rotation and color
- EXIF and GPS
- Removed. A canvas holds pixels and nothing else, so camera model, capture time and location coordinates never reach the PNG. Handy before posting a phone photo in public. If you need those fields, read them first with the metadata viewer.
- Sideways phone photos
- Phones often store the image rotated and add an orientation tag. The browser applies the tag while decoding, so the PNG pixels come out upright and need no tag.
- Wide gamut color
- JPGs tagged Display P3 or Adobe RGB get mapped into sRGB during decoding. Saturated reds and greens lose a little intensity. For print work, keep the original file.
- CMYK JPGs
- Print workflow JPGs in CMYK decode in current browsers, but colors often shift without the press profile. Convert those in the software that made them.
Where this page stops
- 25 files per batch, 50 MB per file. Clear the batch to start another.
- Memory sets the real ceiling. A decoded image takes width times height times 4 bytes, so a 48 MP photo needs about 192 MB before encoding starts. Safari on iPhone caps a canvas near 16.7 megapixels. Bigger files fail at Original size there. Pick 3840 or lower and the file converts.
- Full color output only. No 8 bit palette PNG, no bit depth choice, no interlacing.
- One key color for the whole batch. No per file settings, no brush to restore parts the key removed.
- JPG input only. WebP, GIF and BMP sources go through the format converter. HEIC and AVIF are not handled on either page.
- The ZIP is stored, not compressed. PNG data is already deflated, so compressing the archive again would save almost nothing and cost time.