sxot - Simple X screenshOT

sxot is an extremely simple screenshotting tool for X11 designed with the unix philosophy in mind. This means that by itself sxot doesn’t do much - it simply takes a screenshot and outputs a binary ppm file to stdout. sxot is meant to be used in conjunction with other tools in order to be useful in practice.

Usage

sxot supports only the following options:

• -c, --cursor adds the cursor to the screenshot.
• -g, --geom <x,y,w,h[ wid]> which tells sxot to capture the specified rectangle instead of screenshotting the entire screen. May optionally include a trailing window id.
• -f, --force overwrites the output file if it already exists.
• -w, --window <WID> to capture a specific window.
• -B, --keep-border to capture the window border as well when using -w.

Some example usage:

• Taking a full screen screenshot and saving it to out.ppm:

$ sxot > out.ppm

• Screenshoting a selected area with selx:

$ sxot -g $(selx) > out.ppm

• Screenshoting with a 5 seconds delay with sleep and saving it with a file-name that has date and time information with date:

$ sleep 5 && sxot > $(date '+%F_%T').ppm

• Screenshoting with a 2 seconds delay after selection with slop:

$ sxot --geom $(slop -f "%x,%y,%w,%h"; sleep 2;) > out.ppm

• Saving a png screenshot using ffmpeg:

$ sxot | ffmpeg -i - out.png

Or using ImageMagick:

$ sxot | convert - out.png

• Copying an optimized png screenshot to clipboard, using optipng-pipe and xclip:

$ sxot | optipng-pipe | xclip -selection clipboard -target image/png

• Screenshoting a specific monitor via selx:

$ sxot -g $(selx -m 0) > monitor0.ppm

Output formats

sxot only supports ppm format. Instead of trying to support complicated compressed formats and doing a mediocre job, sxot leaves this task to more specialized tools such as optipng.

Some recommended formats and tools:

• I suggest using png with optipng-pipe for best compatibility and reasonably low file-sizes:

$ sxot | optipng-pipe > out.png

• If smaller file-sizes are desired, then I can suggest using lossless webp with cwebp. Keep in mind that some compatibility is lost since older software might not support webp. Conveniently cwebp accepts stdin if the file is named “-”, although this feature is undocumented.

$ sxot | cwebp -quiet -lossless -z 8 -mt -o out.webp -- -

• For lossy compression, cjpegli is very impressive:

$ sxot | cjpegli /dev/stdin out.jpg -d 0.9 -p 0

The cwebp tool comes with libwebp, and cjpegli is developed alongside the official jpeg-xl library. However, you distro may package these tools separately.

jxl also seems promising in terms of lossless compression. But compatibility isn’t that great as of now (party due to google killing it in chromium).

Dependencies

• Build Dependencies:

  • C11 compiler
  • POSIX 2001 compatible C standard library.
  • Necessary library headers (on some distros you need to install *-dev packages to get header files)

• Runtime Dependencies:

  • X11 server (with true color enabled)
  • Xlib
  • Xfixes
  • libXcomposite (Optional, see the Window capture section)

Building

• Simple release build:

$ cc -o sxot sxot.c -O3 -s -l X11 -l Xfixes

The above command should also work with gcc, clang or any other C compiler that has a POSIX compatible cli interface.

• Debug build with gcc (also works with clang):

$ gcc -o sxot sxot.c -std=c11 -Wall -Wextra -Wpedantic -Wshadow \
    -g3 -D DEBUG -O0 -fsanitize=address,undefined -l X11 -l Xfixes

• If you’re editing the code, you may optionally run some static analysis:

$ make -f etc/analyze.mk

Installing

Simply copy the binary over to your $PATH:

# install -Dm755 sxot /usr/local/bin/sxot

Optionally install the manpage:

# install -Dm644 sxot.1 /usr/local/share/man/man1/sxot.1

A zsh completion script is also available for zsh users under etc/zsh-completion:

# install -Dm644 etc/zsh-completion/_sxot /usr/local/share/zsh/site-functions/_sxot

Window capture

It may seem odd to include -w flag to capture window when tools like selx are already capable of selecting a window and reporting it’s geometry. However there are two cases where simply reporting the window geometry isn’t sufficient:

1. When parts of the window is off-screen. In this case we will not be able to capture the off-screen parts from the root window, since they are off-screen.
2. When a window is being obscured by another window. In this case capturing geometry on the root window will capture the overlapping window instead of only capturing the window that was actually selected.

By using a window id we can capture the window directly instead of the root window which solves both the issues above. Note that both of these cases require a compositor to be running in order to work.

If you have XComposite dynamic library installed then case 1 can sometimes work without a compositor, but it’s not reliable.

The -g flag also optionally accepts a trailing window id, this avoids having to use two different invocation, one for capturing window (-w) and another selection (-g), the following command will work in both cases:

$ sxot -g "$(selx -f "%S" -B)"

Worth noting that the command substitution must be wrapped in double quotes "$(...)", otherwise the shell will split the output. The -B flag to selx is to capture the window border, leave it out if you don’t want it.