I'll be compressing this one 768×768 image (of a closeup of a Romanesco broccoli, taken by me) at a bunch of WebP quality levels, and for comparison the same JPEG quality levels. All the JPEG images here use 4:2:0 chroma subsampling.
Original image
WebP, quality 90
JPEG at quality level 90
WebP, quality 80
JPEG at quality level 80
WebP, quality 50
JPEG at quality level 50
WebP, quality 30
JPEG at quality level 30
WebP, quality 10
JPEG at quality level 10
WebP, quality 0
JPEG at quality level 0
Addendum, December 2022: the difference images shown here are themselves lossy, compressed as WebP's at 80% quality. I have a file size quota to work with here, and when lossless these difference images were over half a megabyte each (or around 750 kB as a PNG), but now they're about 90 kB each, a saving of about 82%.
The difference between a lossless and lossy difference image is noticeable, but at this quality level the difference seems to mostly be in vibrancy or chroma: the location of the differing pixels is correct, but the dark greens of the original difference image are more akin to grey when compressed. The reds and blues survived compression a bit better than the greens. Also, in this case, the differences are mostly individual pixels, rather than large areas of color (like seen in the lower-quality JPEG compressions); large areas of color survive compression better. The greying-out of these images isn't a purely WebP thing either: JPEG does the same thing. The contrast difference between a single green pixel between two black pixels seems to just be too great.
In any case, the original image still exists, so lossless difference pictures can still be computed.