Quality 80 does not mean 80% of the image. It is a setting each encoder interprets for itself — and on a PNG, it is ignored entirely.
What lossy actually discards
JPEG and lossy WebP throw away high-frequency detail: fine texture, film grain, small color transitions. At normal viewing size your eye skips most of it, so the encoder spends bytes on the structure you do notice and drops the rest. PNG, by contrast, stores every pixel exactly.
| Lossless (PNG) | Lossy (JPEG, WebP) | |
|---|---|---|
| Pixel values after saving | Exact | Approximate |
| Repeated saves | No cumulative loss | Degrades each time |
| Transparency | Yes | No in JPEG; yes in WebP |
| Typical use | Screenshots, logos, UI | Photographs |
| Size for photos | Large | Much smaller |
Why quality 80 usually looks fine
Human vision is forgiving of high-frequency loss, and screen viewing hides more of it than print. Quality 80 keeps shapes, edges, and most texture; the last stretch from 85 to 100 adds bytes much faster than visible detail. The exception is text: a screenshot of small type at quality 80 can pick up ringing around the letters, which is why screenshots belong in PNG.
The trap in the compressor
The image compressor keeps whatever format you feed it and applies the slider — 1 to 100, default 80. JPEG and WebP inputs get re-encoded and shrink. A PNG input stays PNG, and the canvas PNG encoder ignores the quality argument, so dragging the slider from 20 to 100 will not meaningfully change the file size. On browsers too old to encode WebP, a WebP input falls back to PNG output.
For real savings on a PNG, convert it to WebP with the format converter; that step is a lossy encode at fixed quality 0.92. Each conversion compounds the damage, so convert from the original, not from an already compressed copy.
Keep the original, compress a copy once, and never re-compress a compressed photo — loss accumulates, and the second pass buys nothing but artifacts.