9 Of 674 Gallery JPEGs Are Progressive. The Rest Are Baseline

We keep telling people progressive is free. Then we counted. We are not using it.

Published September 19, 2026 · RealAIGirls · About a 4 minute read

We keep writing about progressive JPEG as if we use it. Today we opened the 674 JPEG files sitting in the gallery dump and asked Pillow one question: is this file progressive, or is it baseline?

9 files are progressive. 665 are baseline. That is 1.3 percent. The rest of the site is still shipping the older scan pattern, the one that has to finish the whole file before a browser can paint a full looking preview.

What we counted

The dump is 674 files with a .jpg or .jpeg name. Pillow reports a progressive flag on the JPEG decoder. We did not reencode anything. We did not throw out truncated files. Every file opened. Mean size was 354,170 bytes.

The 9 progressive files are not a secret setting we turned on last week. They are leftovers. The other 665 were saved the default way and never revisited.

Why this number matters

Yesterday we measured what progressive does when you actually turn it on: about 3 percent fewer bytes at quality 85 on 400 pictures, with a win on most of the set. That is a real saving. It is also a setting we have almost never applied to the files people download.

A compression guide that does not look at the live files is a lecture. This is the live files. We told people progressive was free. We then served baseline on 665 of 674 JPEGs.

What we are not claiming

This is not a quality test. We did not score SSIM or PSNR. We did not prove the 9 progressive files look better or worse. We counted a flag. If we reencode the 665, the byte math from yesterday is the prediction, not a second measurement. That work is next, and it should be done on these exact files, not on a fresh sample we cherry pick because it behaves.