# JPEG vs PNG: We Measured Both, Here Is When Each One Wins

> JPEG vs PNG with real numbers: the same photo came out 6x larger as PNG, the same graphic 25x larger as JPEG. Which format wins for photos, screenshots, logos and uploads.

*Published: 2026-08-27* · *6 min read*

Canonical URL: https://timestampcamera.net/photo-guides/jpeg-vs-png


**Quick answer:** Use **JPEG for photographs** and **PNG for graphics, screenshots with text, logos, and anything transparent**. This is not a matter of taste; the formats are built for opposite kinds of images, and picking the wrong one costs real bytes or real quality. We measured it: the same photograph saved as PNG came out **6 times larger** than the JPEG with no visible benefit, while the same flat-color graphic saved as JPEG came out **25 times larger** than the PNG and picked up visible edge artifacts on top. The rest of this piece is those numbers, the two or three real exceptions, and the metadata difference nobody mentions.

## The experiment

Format debates attract a lot of "it depends" hand-waving, so we ran the comparison instead of describing it. Two 1200 x 800 images, each saved both ways with the same settings (JPEG at quality 85, PNG with standard optimization):

**Image A, photo-like:** a smooth gradient with per-pixel noise, which is what real camera output looks like up close, since every sensor adds noise.

**Image B, graphic-like:** four flat color blocks with clean white lines across them, which is what charts, screenshots of interfaces, and logos look like up close.

| Same pixels, both formats | JPEG (85) | PNG | Verdict |
| --- | --- | --- | --- |
| Photo-like image | **239 KB** | 1,428 KB | PNG is 6.0x larger |
| Graphic-like image | 112 KB | **4.7 KB** | JPEG is 25x larger |

Same machine, same dimensions, same minute. The only variable is what kind of image went in, and the result flips completely.

And the graphic did not just cost more as a JPEG; it got worse. We sampled a pixel right at one of the hard color edges in the JPEG version and got back a value of (244, 255, 255) where the source was pure white. That off-white smudge is JPEG's edge ringing, the faint halo you have seen around text in over-compressed screenshots. The PNG's edge pixel was exactly what went in.

## Why the same format wins by 6x and loses by 25x

The two formats make opposite bets about what an image is.

**JPEG bets the image is a photo.** It carves the picture into 8 x 8 blocks and discards fine detail the eye is bad at noticing, especially subtle variations in smooth areas. Sensor noise and soft gradients are exactly that kind of detail, so photos shrink enormously with little visible cost. Sharp edges are the opposite of what it expects, and it smears them.

**PNG bets nothing is disposable.** It compresses losslessly, finding repetition instead of discarding detail. Flat color regions are pure repetition, so a chart packs down to almost nothing. Per-pixel noise is the opposite of repetition, which is why our noisy photo ballooned to 1,428 KB: PNG faithfully stored every speck of grain.

That is the whole story. Photos are noise and gradients, so JPEG wins. Graphics are edges and flat fills, so PNG wins. Screenshots sit with graphics almost always, the exception being a screenshot that is mostly a photo.

## The two capabilities that decide edge cases

When the size logic does not settle it, one of these usually does.

**Transparency.** PNG carries a real alpha channel; JPEG has nowhere to put one. Save a logo with a transparent background as JPEG and the transparency is gone, flattened onto white. This is not subtle and there is no workaround inside JPEG; if you need transparency, the decision is made for you. It is also why our [PNG to JPG converter](/png-to-jpg) asks what background color to flatten onto, rather than pretending the question does not exist.

**Metadata.** JPEGs carry EXIF: capture date, camera, GPS, the lot. PNG technically gained an eXIf chunk in a later spec revision, but in practice almost nothing writes it, and browser conversions drop metadata wholesale; we dug into the details in [Do PNG files have EXIF data?](/photo-guides/do-png-files-have-exif-data). This cuts both ways. Converting a JPEG to PNG usually silently discards the photo's history, which is a loss if you care when and where it was taken, and a quiet privacy win if you were about to post it. Either way it should be a choice, not an accident: check what a file carries with the [EXIF viewer](/exif-viewer), and strip deliberately with the [EXIF remover](/exif-remover) rather than laundering metadata through a format change.

## Practical calls, one line each

- **Photo for email, web, or a form upload:** JPEG, quality 80 to 90. If there is a size cap, aim at it with the [image compressor](/image-compressor).
- **Screenshot of an app or website:** PNG. Text stays crisp and the file stays small.
- **Logo or icon:** PNG, for the edges and the transparency.
- **Chart or diagram exported from a tool:** PNG, by a mile; our test says 25x.
- **Scanned document:** JPEG if it is essentially a photo of paper; PNG if it is high-contrast text you may need to read closely or OCR.
- **Photo you plan to edit repeatedly:** PNG or a lossless original, because every JPEG re-save discards a little more. Edit once, export JPEG last.
- **Need the smallest possible modern file:** neither; WebP and AVIF beat both, with compatibility trade-offs we covered in [WebP vs AVIF vs JXL](/photo-guides/webp-vs-avif-vs-jxl).

## Converting between them

Both directions are legitimate and both run in your browser without uploading anything: [JPG to PNG](/jpg-to-png) when you need a lossless copy or are about to edit, [PNG to JPG](/png-to-jpg) when a portal insists on JPEG or the PNG is absurdly large for what it shows.

One conversion myth worth killing: converting a JPEG to PNG does not recover quality. The detail JPEG discarded is gone; the PNG just preserves the damage losslessly, in a much larger file. Conversion changes the container, never the history.

## Frequently asked questions

**Is JPEG or PNG better quality?**
For identical pixels, PNG is technically perfect because it is lossless, while JPEG throws away detail to save space. On a photograph the difference is usually invisible at quality 85 and above. On sharp-edged graphics JPEG shows halos around edges, which is where PNG's losslessness earns its keep.

**Which is smaller, JPEG or PNG?**
Entirely content-dependent, as the table above shows: the photograph was 239 KB as JPEG and 1,428 KB as PNG, while the flat-color chart was 112 KB as JPEG and only 4.7 KB as PNG. Photos compress smaller as JPEG; graphics compress smaller as PNG.

**Does PNG support transparency?**
Yes, a full alpha channel. JPEG has none, so anything transparent gets flattened onto a background color the moment you save as JPEG.

**Do PNG files keep EXIF data?**
Mostly no in practice. The spec added an eXIf chunk, but very little software writes it, and browser conversions drop metadata entirely. If a photo's capture date or GPS matters to you, keep the JPEG original.

## Bottom line

Photograph, JPEG. Graphic, screenshot, logo, or anything transparent, PNG. Our own measurements put the penalty for guessing wrong at 6x wasted space in one direction and 25x plus visible artifacts in the other, so the guess is worth getting right. And when a photo changes format, remember its metadata probably did not survive the trip: verify, do not assume.