A screenshot saved as PNG can easily reach 2–5 MB, while the same image saved as JPEG comes in at around 200 KB — a 10–25× size difference for visually identical results. This is the trade-off at the heart of the PNG format: lossless compression that stores every pixel precisely, with no approximation or quality degradation. Understanding why PNGs are large is the first step toward making them smaller without sacrificing what matters.

Why PNG Files Are Large

PNG was designed for lossless storage. Every pixel value is recorded exactly, which means the compression algorithm must reconstruct the original image bit-for-bit from the compressed data. This precision comes with inherent size costs:

Full pixel fidelity — A 1920×1080 image with 32-bit color (RGBA) has over 8 MB of raw pixel data. PNG's Deflate compression reduces this, but cannot approach JPEG's lossy ratios.
Deflate compression ceiling — PNG uses the same Deflate algorithm as ZIP files. Lossless compression only exploits redundancy — repeated patterns, flat color regions. Complex photographic content has little redundancy to compress.
Alpha channel overhead — Transparency (the fourth RGBA channel) adds a full 8 bits per pixel. A 1920×1080 RGBA image carries ~2 MB of alpha data alone.
Embedded metadata — PNG files can contain EXIF data, ICC color profiles, creation timestamps, and software comments — often adding 5–100 KB that serves no display purpose.

Lossless PNG Compression

Lossless compression tools re-encode the PNG file using more aggressive Deflate settings, optimized filter selection, and metadata stripping — without changing any pixel values. The output is byte-for-byte identical to the input when decoded. Typical savings: 10–30%.

Popular lossless PNG optimizers:

oxipng — Fast, multi-threaded Rust implementation. Tries multiple compression strategies and picks the best. Recommended for batch processing.
pngcrush — The classic tool. Tries 114 different compression methods and uses the smallest result. Slower than oxipng but widely supported.
Zopfli — Google's ultra-compression library. Produces the smallest possible Deflate-compressed output, but is 100× slower than standard tools. Ideal for one-time optimization of static assets that won't change.

These tools are safe for any PNG — the result is visually and numerically identical to the original. They are the right choice when pixel-perfect accuracy is required: medical images, design source files, screenshots for documentation.

Lossy PNG Compression

Lossy PNG compression applies quantization to reduce the number of unique colors in the image before encoding. The most widely used technique — popularized by pngquant — converts a 32-bit RGBA PNG (up to 16.7 million colors) to an 8-bit indexed-color PNG (up to 256 colors). Typical savings: 50–80%.

This sounds drastic, but intelligent dithering algorithms scatter color approximation errors across neighboring pixels so that the human eye perceives a smooth result. For logos, icons, illustrations, UI screenshots, and graphics with large flat-color regions, the visual difference is negligible. For photographs or images with subtle gradients, slight color banding may be visible at high reduction ratios.

Best for lossy PNG: logos, icons, app screenshots, web UI graphics, illustrations with flat colors.
Avoid lossy PNG for: photographs, images with gradients, source files that will be edited further.

Color Depth Reduction

PNG supports several color modes. Choosing the right one for your image is a simple way to achieve large file size reductions without any perceptible quality change:

8-bit indexed (palette) PNG — Up to 256 colors. Best for logos, icons, pixel art, illustrations, and any image with flat color regions. Files can be 60–80% smaller than 24-bit equivalents. Most design tools export PNGs as 32-bit by default even when the image only uses 32 colors.
24-bit RGB PNG — 16.7 million colors, no transparency. Best for photographs and images with complex color gradients where the full color space is needed. Smaller than 32-bit because there is no alpha channel.
32-bit RGBA PNG — 16.7 million colors plus transparency. Required only when the image has actual transparency. Many exported PNGs include an alpha channel even when every pixel is fully opaque — removing this drops file size by ~25%.

When evaluating a PNG, ask: does this image actually use transparency? Does it use more than 256 distinct colors? If the answers are no and no, an 8-bit palette PNG is appropriate and will be dramatically smaller.

Alpha Channel Optimization

The alpha channel is the most frequently over-used feature in PNG files. Before keeping a 32-bit RGBA PNG, consider:

No transparency needed? — Convert RGBA to RGB. This removes the alpha channel entirely, reducing raw pixel data by 25% before compression. For web images placed on solid-color backgrounds, this is almost always the right call.
Binary transparency? — If pixels are either fully opaque or fully transparent (no partial transparency), consider using PNG's tRNS transparency chunk instead of a full alpha channel. This stores transparency as a palette index rather than per-pixel values.
Simplify semi-transparent edges — Anti-aliased edges against a known background color can be pre-composited, eliminating the need for alpha entirely. A logo on a white card, for example, can be exported as RGB rather than RGBA.
No transparency at all? — Convert to JPG. If a PNG has no transparency and contains a photograph, JPEG will be dramatically smaller with equivalent visual quality.

When to Switch from PNG to WebP or AVIF

For web use, the most impactful optimization is often switching formats entirely. Modern image formats achieve better compression than PNG at the same visual quality:

WebP lossless — Google's WebP format in lossless mode is on average 26% smaller than PNG for the same image. It also supports transparency. Supported by all modern browsers. For web graphics, converting PNG to WebP lossless is almost always a win.
WebP lossy — For photographs that were mistakenly saved as PNG, WebP lossy achieves 25–34% smaller files than JPEG at equivalent visual quality. Much better than trying to compress a photo PNG.
AVIF — The newest format. AVIF lossless can be 50% smaller than PNG. Excellent for cutting-edge optimization. Browser support is strong (Chrome, Firefox, Safari 16+), though slightly narrower than WebP.

Use Picovert's image converter to convert PNG files to WebP or AVIF without installing any software. The conversion runs entirely in your browser.

Online PNG Compressor

For quick PNG compression without installing software, Picovert's image compressor handles PNG files directly in your browser. Drop one or multiple PNG files, and download the compressed versions instantly. No account required, no files uploaded to any server — all processing happens locally.

The compressor applies lossless optimization: metadata stripping, optimized Deflate re-encoding, and unused chunk removal. For most PNG graphics, this produces 15–40% smaller files with zero visible quality change.

Command-Line Tools

For automated workflows and batch processing, command-line tools offer fine-grained control:

pngquant (lossy, 50–80% reduction):
pngquant --quality=65-80 *.png
Processes all PNGs in the current directory. The --quality flag sets the acceptable quality range (0–100). Lower values produce smaller files with more color reduction. Output files are named original-fs8.png by default.
oxipng (lossless, 10–30% reduction):
oxipng -o 6 *.png
The -o 6 flag sets optimization level (0–6, where 6 is maximum compression). Higher levels take longer but produce smaller files. Use --strip all to also remove metadata.
Combine both: Run pngquant first (lossy quantization), then oxipng (lossless re-compression of the quantized output). This stacks both savings and typically achieves 60–75% total reduction.

PNG Compression Decision Tree

Not sure which technique to use? Follow this decision process:

Does the image contain a photograph or complex gradients? — Yes: convert to JPEG (no transparency needed) or WebP. PNG is the wrong format for photographs. No: continue.
Is this for the web? — Yes: convert to WebP lossless (or AVIF for maximum compression). Both support transparency and compress better than PNG. No: continue.
Does the image use transparency? — No: convert RGBA to RGB (or JPEG if it's a photo). Yes: continue.
Does the image use more than 256 colors? — No: use 8-bit palette PNG (lossy quantization with pngquant). Yes: use 24-bit or 32-bit PNG with lossless optimization.
For pixel-perfect accuracy — Use lossless optimization (oxipng, pngcrush) and strip metadata. Expect 10–30% savings.

For most web and sharing use cases, start with Picovert's free PNG compressor — it handles lossless optimization automatically and delivers results in seconds, with no software to install. When you need deeper savings or format conversion, the image converter and the command-line recipes above have you covered.