Image Tools

PNG Compressor

Lossless PNG re-encoding with honest results per file — and clear guidance when a format switch would save far more.

In short

Compress PNG images online losslessly — identical pixels, smaller files where possible, with honest per-image savings shown. Free and private.

By NaveenKumar T · Updated 18 Sep 2026

Learn: PNG vs JPG: The Differences That Actually Matter

Per-image savings shown
3 quality presets or a KB target
Batch compression

Runs entirely in your browser — nothing you paste or open here is uploaded, logged, or stored.How we handle data →

Learn more

Compressing PNG images, explained

PNG compression here is lossless and runs on oxipng (the modern Rust successor to OptiPNG, compiled to WebAssembly): your pixels come out identical, and savings come from re-packing the data more efficiently than the original encoder did. Files exported by design tools, games, and older software often carry slack that oxipng reclaims; PNGs that were already optimized will report 'already optimized' rather than pretending — this tool never inflates a file to fake a win.

The honest big-picture: when a PNG is photographic (a screenshot of a photo, a camera image saved as PNG), no lossless trick competes with changing format — the same image as WebP or high-quality JPG is typically 5–10× smaller. That's why the Related tools below include the converters; the right compression is sometimes a conversion.

Which PNGs shrink, and by how much

The source matters more than the preset. Figma, Sketch and Illustrator exports commonly lose 10–30%, since design tools optimize for speed rather than size; macOS and Windows screenshots 10–25%; images from older or minimal encoders (MS Paint, GDI+, some game engines) 20–50%. Files that already went through pngcrush, ImageOptim or TinyPNG report 0–2% and come back untouched. Savings come from lossless tricks oxipng applies to the original bytes without ever decoding to a canvas: better filter selection, tighter Deflate, and reductions like dropping an alpha channel that is fully opaque or a 16-bit depth that only holds 8-bit values. Transparency is preserved exactly.

The reductions that make lossy PNG tools famous — quantizing to 256 colors, which is how TinyPNG-style services reach 60–80% — are deliberately not done here, because they alter pixels. When a PNG is a photograph, a screenshot of a photo, or artwork with soft gradients, the honest route to big savings is a format change: PNG to WebP keeps the alpha channel and typically lands 3–5× smaller, PNG to AVIF smaller still. Run this compressor first across an assets folder as safe, review-free housekeeping — identical pixels need no visual QA — then convert the photographic stragglers. Outputs are named name-compressed.png; strip the suffix to overwrite in place.

Help

Frequently asked questions

Everything you need to know about the PNG Compressor.

01Is PNG compression here lossless — do any pixels change?

PNG compression here is fully lossless — no pixels change. oxipng never decodes the image to a canvas; it rewrites the original bytes with better scan-line filters, tighter Deflate and lossless reductions such as dropping a fully opaque alpha channel or a 16-bit depth that only holds 8-bit values. Transparency is preserved exactly. The 60–80% savings that lossy PNG services advertise come from quantising to 256 colours, which this page deliberately does not do.

02Why did my PNG only shrink a few percent?

It was probably already well-encoded — lossless formats have a hard floor, and hitting it means there was nothing left to reclaim. If the image doesn't need transparency or pixel-perfect edges, converting to WebP (lossless or lossy) is the lever that actually moves: try PNG to WebP from Related tools.

03What do the three presets change if PNG is lossless?

Effort, not quality. High Quality runs oxipng at optimization level 2, Balanced at level 3, Max Compression at level 4 — each tries more combinations of scan-line filters and Deflate strategies to find a smaller encoding of the exact same pixels. The gain from level 2 to 4 is usually only 1–3% for several times the processing time, so Balanced is the sensible default for large batches and Max is worth it only when every kilobyte counts, such as icons shipped in an app bundle.

Keep working