PNG is the go-to format for logos, screenshots, and any graphic that needs a transparent background - but that same quality comes at a cost. A truecolor PNG can store up to 16.7 million colors, and most images only use a fraction of that palette. A PNG compressor exploits exactly that gap, cutting file size dramatically without touching the parts of the image anyone would actually notice.

This guide covers what actually happens inside a PNG compressor, why "lossy" and "PNG" aren't contradictory terms, and how to compress a PNG properly depending on what the image is for.

How Does PNG Compression Work?

PNG files are compressed in one of two fundamentally different ways, and knowing which one a tool uses explains why results vary so much between compressors.

Lossless re-encoding is what happens by default when a PNG is saved. It uses the Deflate algorithm - a combination of LZ77 pattern-matching and Huffman coding - to eliminate redundant data without touching a single pixel. A well-optimized PNG compressor can still squeeze another 10-30% out of an already-saved file, since most image editors don't apply the most efficient Deflate settings by default.

Color quantization is where the bigger gains happen, and it's the technique behind tools like TinyPNG and pngquant. Instead of storing every one of an image's millions of possible colors, the compressor analyzes the actual colors used, builds an optimized palette of 256 colors or fewer, and maps each pixel to its nearest match - often with dithering applied so the transition looks smooth rather than banded. Because a PNG can store the same image in an 8-bit indexed format instead of 24-bit truecolor, this alone typically cuts file size by 60-80%.

Is PNG Compression Lossy or Lossless?

Both options exist, and the right one depends entirely on what the image contains.

• Lossless compression keeps every pixel identical and is the safer default for source files, technical diagrams, or anything you plan to re-edit later. Expect modest savings of roughly 10-30%.
• Lossy compression (quantization) reduces the color count and typically delivers 60-80% smaller files. For most real-world PNGs - logos, UI screenshots, icons with flat color areas - the visual difference is imperceptible at normal zoom levels.
• Where quantization struggles: photographic PNGs and images with smooth gradients can show subtle banding once the palette drops below roughly 128 colors, since a gradient depends on thousands of intermediate shades that get merged together.

Does PNG Compression Keep Transparency?

Yes. Both lossless re-encoding and color quantization fully preserve the alpha channel. PNG-8 (indexed) format stores a transparency value per palette entry, while PNG-24/32 keeps a full 8-bit alpha channel per pixel - either way, a properly compressed PNG will still work as a transparent overlay, app icon, or CSS background exactly as before.

How to Compress a PNG Step by Step

1. Identify what the PNG is for. A logo or icon with flat colors can drop to a 32-64 color palette with essentially no visible change. A screenshot with anti-aliased text needs a higher color count (128-256) to avoid banding around the letters. A source file you'll re-edit later should stay lossless.
2. Choose quantization for the biggest savings. If the image doesn't need pixel-perfect precision, a quantization-based tool (TinyPNG, pngquant, or any compressor built on the same median-cut algorithm) will outperform a purely lossless re-encode by a wide margin.
3. Preview before publishing. Nearly every PNG compressor worth using shows a before/after comparison. Zoom in on text edges and gradients specifically - that's where quantization artifacts show up first, if they show up at all.
4. Batch it if you have more than a handful. Compressing icon sets or full screenshot libraries one at a time wastes time; look for a tool that accepts multiple files and returns a zipped batch.
5. Consider whether PNG is even the right format. If the image doesn't need transparency, converting it to JPEG or WebP first will almost always produce a smaller file than any amount of PNG-specific compression - PNG's overhead exists specifically to support things JPEG can't do.

When Should You Avoid Compressing a PNG?

Not every PNG benefits from aggressive compression, and knowing when to leave a file alone matters as much as knowing how to shrink one.

• Pixel art, where the palette is already tiny and every pixel is deliberate - quantization has nothing left to gain here and can distort intentional color choices.
• Archival or source files you expect to re-edit later, where preserving full color depth avoids compounding quality loss across multiple save cycles.
• Technical diagrams or medical/scientific imagery, where exact color values may carry meaning that quantization would blur.

Conclusion

A PNG compressor doesn't have to mean a trade-off between file size and quality - color quantization exploits the simple fact that most images use far fewer colors than the format can technically store. For everyday logos, screenshots, and web graphics, lossy quantization down to 256 colors or fewer usually delivers 60-80% smaller files with no visible difference. Reserve lossless compression for source files, pixel art, and anything where exact color values matter, and always preview the result before it goes live.