pngquant

pngquant is a PNG compressor that reduces file sizes by converting images to an efficient 8-bit PNG format with an alpha channel. This typically results in files that are 60-80% smaller than 24/32-bit PNGs while remaining standards-compliant.

Basic Usage

• Batch conversion: Process multiple files at once using wildcards.
• Stdin/Stdout: Chain with other Unix tools using pngquant -.

To achieve even smaller file sizes, it is recommended to use pngquant in conjunction with tools like oxipng, ImageOptim, or zopflipng.

Dithering Control

• --nofs: Disables Floyd-Steinberg dithering entirely.
• --floyd=0.5: Controls the level of dithering where 0 is none and 1 is full. Note that the = character is required (e.g., --floyd=0.5).

Color Precision (--posterize bits)

Reduces the precision of the palette by a specified number of bits. This is useful when images are intended for low-depth screens (e.g., 16-bit displays or ARGB444 textures).

Set a quality range using two numbers from 0 (worst) to 100 (perfect).

• pngquant will use the minimum number of colors required to meet or exceed the max quality value.
• If the resulting quality is below the min value, the image will not be saved (if outputting to stdout, the 24-bit original is returned) and the program exits with status code 99.

Example: pngquant --quality=65-80 image.png

Custom Output Path (-o or --output)

Specify a specific path for the converted file. Note that when using this option, you are restricted to a single input file.

Custom File Extension (--ext)

Set a custom suffix for the output filename. By default, pngquant uses -or8.png or -fs8.png.

Warning: If you use --ext=.png --force, pngquant will overwrite the original input files in place. Use this with caution.

Speed/Quality Trade-off (--speed N)

Adjust the balance between speed and compression quality using a value from 1 to 11:

• 1: Slowest, highest quality, smallest files.
• 11: Fastest, less consistent quality, lighter compression.
• Default: 4. It is recommended to keep the default unless generating images in real-time (e.g., map tiles).

Skip larger files (--skip-if-larger)

Prevents writing the converted file if the compression does not actually reduce the file size.

Strip metadata (--strip)

Prevents the copying of optional PNG chunks (metadata).

Adjust the compression speed using a value between 1 (slowest) and 11 (fastest).

• Values 1-9: Standard speed range.
• Value 10: Enables fast compression.
• Value 11: Enables fast compression and disables dithering (--floyd=0.0).

You can specify how the output files are named using either --output or --ext. Note that these two options cannot be used at the same time.

• --output <file>: Specifies a single output file. When using this, only one input file is allowed. Use - to write to stdout.
• --ext <extension>: Specifies a new file extension to append to the original filename (e.g., --ext .webp). If not provided, pngquant defaults to -fs8.png (if Floyd-Steinberg dithering is used) or -or8.png (if ordered dithering is used).

Use these options to reduce the color palette or simplify the image structure:

• --colors <N>: Limits the number of colors in the output image. Must be between 2 and 256.
• --posterize <bits>: Reduces the number of bits per color channel. Must be between 0 and 4 bits.

Set the target quality range for the lossy compression. The quality values are numbers between 0 (lousy) and 100 (perfect).

Supported formats:

• N-M: No worse than N, no better than M (e.g., 60-80).
• N-: No worse than N, perfect if possible (same as N-100).
• -N: No better than N (same as 0-N).
• N: Target quality N (internally sets range to N*0.9 to N).