Two different jobs that look like one
Deciding how far a hero image can be compressed before it visibly degrades is a careful, visual, one-image job, and Squoosh is built precisely for it: two panes, a draggable divider, every encoder parameter exposed. If that is your task, use it, and this page will not pretend otherwise.
Converting the sixty PNGs in a content folder to WebP before a site launch is a completely different job. It needs no visual comparison and no per-image decision - it needs one setting applied sixty times without sixty separate interactions. Doing that in a single-image tool is an hour of clicking.
What PNG to WebP actually buys you
WebP in lossless mode typically produces files noticeably smaller than an equivalent PNG, with the alpha channel intact - so a logo or a screenshot converts with no visible change and less weight. In lossy mode, with transparency still supported, the saving on screenshots and illustrations is much larger again.
That matters because images are usually most of a page weight, and the largest one is usually what the loading metrics are measuring. Converting a folder of PNG screenshots to WebP is one of the few optimisations that is genuinely free: nothing about the page changes except how long it takes to arrive.
Resize before you re-encode
The biggest saving is almost never the codec. An image exported at 3000 pixels wide and displayed in a 700 pixel column is carrying roughly twenty times the pixels it needs, and no encoder recovers that - it just compresses the waste efficiently.
Set the output width to the size the image is actually displayed at, or twice that for high-density screens, and then convert. Doing both in one pass is the reason width sits next to format here: the resize is the part that matters and it is the part people skip.