Where the savings come from
Most images on disk are saved at a higher quality setting than anyone can see. Cameras export JPGs at 95+, design tools default to "maximum", and the result is files two to five times larger than they need to be. Re-encoding at quality 70–80 keeps the visible image and discards the bytes spent on imperceptible detail.
That is all this tool does — honestly. There is no magic "50% smaller, zero loss" claim: lossy compression always trades microscopic fidelity for size. The point is choosing that trade deliberately, with a live before/after readout and a full-size Compare view on every file.