SIZEMINT GUIDES
PNG has two ways to get smaller: lossless optimization that changes nothing, and palette quantization that looks identical on graphics. Here is how both work and when to use each.
PNG stores every pixel exactly — that is its job. For a logo with flat colors, "exactly" costs little. For a photograph with millions of subtly different pixels, "exactly" can cost 5–10MB where a JPG would need 300KB. Before optimizing a PNG, ask whether the image should be a PNG at all: photographs almost never should be. If it is a photo, convert it to WebP or JPG first — that single decision dwarfs every optimization below.
Lossless optimizers rewrite the PNG's internal encoding more efficiently — better filtering, better deflate settings, stripped metadata — without touching a single pixel. Think of it as zipping the file smarter. Typical savings are 10–40%, the image is bit-for-bit identical, and there is no reason not to do it. This is always the first step.
Quantization reduces the color palette — say, from 16 million colors to 256. That sounds drastic, but a logo or UI screenshot rarely uses more than a few hundred distinct colors, so the result is visually identical while the file can shrink 60–80%. This is technically lossy, which is why the method matters: use it on graphics, icons, charts and screenshots — never on photographs or gradients, where the missing colors show up as visible banding.
Use PNG when you need transparency (logos, overlays, watermarks), pixel-perfect screenshots for documentation, or line art where JPG's smudging around sharp edges would show. For everything else, pick the format by content: photos go to JPG or WebP, simple graphics stay PNG and get quantized. Choosing the right format up front beats any amount of squeezing the wrong one.
Open the SizeMint compressor and drop in your PNG. Processing happens entirely in your browser — the file is never uploaded.
Enter the KB target you need. The compressor searches encodings automatically for the best result that fits, instead of making you guess a quality number.
Zoom to 100% and inspect sharp edges, small text and transparency boundaries — these are where over-compression shows first. If edges look dirty, the target is too aggressive for the image's complexity: raise it or reduce the image dimensions first.
If your PNG has a transparent background, check it against a dark backdrop after compression. A flattened or fringed edge means the transparency was mishandled — reconvert from the original.
A logo or icon: lossless optimization alone often cuts 20–40%; with quantization, 60–80% smaller at visually identical quality is normal. A full-screen screenshot: expect 40–70% with careful settings. A photograph saved as PNG: convert it to JPG or WebP instead — no PNG technique will beat simply using the right format, and you will typically land at one-tenth the size.
Lossless optimization never does — the pixels are identical. Quantization is technically lossy but visually lossless on graphics with limited colors. On photographs, both the wrong method and visible damage.
Almost always because it is a photograph (or a very large image) in a format built for exactness. Convert photos to JPG or WebP. If it is a graphic, check the dimensions — a 4000px-wide logo carries 16× the pixels of a 1000px one.
PNG for documentation and anything with text — JPG smudges sharp edges. If the screenshot is mostly a photograph (a video still, a game capture), JPG or WebP at high quality is fine and far smaller.
Yes — transparent pixels stay transparent through compression. Verify against a dark background after downloading, as described above.
Choosing between modern formats? Read WebP vs JPG: which is smaller. For exact upload limits, see how to compress an image to 20KB.
Have a PNG to shrink? Open the SizeMint compressor, set your exact target, and keep every pixel.
