New AI Trick Shrinks Image Files 10% Without Losing Quality

A team of researchers has unveiled a new way to shrink digital images without hurting how they look. Their system, called MIRC, uses artificial intelligence to squeeze photos and video frames into smaller files. In tests, it produced files roughly 10% smaller than VVC — one of the most efficient compression standards used today — while keeping the same image quality.

Traditional compression works by spotting repetitive patterns and removing redundant data. MIRC goes further by tailoring itself to each individual image. It studies the picture, learns its details, and then builds a custom compression recipe just for that file. This "overfitted" approach has been tried before, but earlier versions were slow and inefficient. MIRC fixes that by doing all the heavy lifting upfront, then decompressing very quickly.

Why does this matter to you? Smaller image files mean web pages load faster, video calls use less bandwidth, and your phone's storage lasts longer. Streaming services could deliver higher quality video without using more data. Apps, photo backups, and cloud services would also become cheaper to run — savings that might eventually reach your monthly bill.

The catch: the compression step takes extra computing power compared to standard tools. That's fine for pre-recorded videos or photos uploaded to the cloud, but it's not yet practical for live video chats on a cheap phone. Still, the team says the decoding side is light enough for many devices, and they've made configurations that let manufacturers choose their preferred balance between speed and file size.