What does "lossless" PNG compression mean?
PNG is already a lossless format — every original pixel survives a decode intact. What the oxipng and libspng encoders do is find a smaller way to express the same pixel data. They try different filter strategies per row, multiple zlib compression levels, and palette reductions for images with 256 or fewer unique colors. The output decodes to the exact same pixel grid, but the file is smaller.
This is the safe option when you cannot afford any quality change: UI screenshots, logos, icons, transparent overlays, technical diagrams, and assets that get re-encoded by other tools later.
How much can a PNG shrink?
It depends entirely on the image. A flat-color logo with transparency can drop 50–80%. A photograph saved as PNG usually shrinks only 5–15%, because the noise in photographs defeats palette-based tricks — for those, re-encoding to JPEG or WebP will do far more. Pixshrink shows you the byte delta before you download, so you can decide whether the savings justify the recompression.
For UI assets with sharp edges and a small palette, the preset = max option pushes the encoder through more trials, usually adding 10–20% extra savings on top of the default pass.
What about transparency?
Alpha channels are preserved exactly. PNG's 8-bit alpha survives the round-trip, and Pixshrink does not "premultiply" anything that would change how your image composites against its background. If your image has a 16-bit per channel color depth, that is preserved too — you will not silently get downgraded to 8-bit.
When to use a different format instead
If you control the file's destination, consider converting PNG to WebP for 25–35% additional savings at the same quality, or to AVIF for 40–50%. For animated PNGs, GIF is no longer the only option — most browsers now play animated WebP, and Pixshrink will pick a smaller output automatically.