Skip to content
kootrit edited this page Nov 7, 2024 · 45 revisions

Image Previews

With Überzug

Note

The original Überzug project is no longer maintained. You might want to use one of the alternative forks instead:

If you want to use Überzug for image previews, it will be necessary to start both lf and Überzug in a wrapper script. Place a file with the following contents in your PATH:

#!/bin/sh
set -euf

if [ -n "${DISPLAY-}" ] && [ -z "${FIFO_UEBERZUG-}" ]; then
  export FIFO_UEBERZUG="${TMPDIR:-/tmp}/lf-ueberzug-$$"

  cleanup() {
    exec 3>&-
    rm -- "$FIFO_UEBERZUG"
  }

  mkfifo -- "$FIFO_UEBERZUG"
  # Execute ueberzug in a loop in case it crashes. Ueberzug dies if its
  # associated window is closed. This breaks image previews when using tmux and
  # reattaching to an existing session.
  while [ -p "$FIFO_UEBERZUG" ] && ! ueberzug layer -s <"$FIFO_UEBERZUG"; do :; done &
  # Open the FIFO for writing. FIFO readers receive an EOF once all writers
  # have closed their respective file descriptors. Holding a file descriptor
  # will effectively keep ueberzug alive as long as lf lives.
  exec 3>"$FIFO_UEBERZUG"
  trap cleanup EXIT

  # Start lf without passing in the file descriptor. This is done to avoid the
  # lf server being passed the file descriptor, which would cause ueberzug to
  # live longer than is strictly necessary.
  lf "$@" 3>&-
else
  exec lf "$@"
fi

Now, make a file named cleaner in the lf configuration directory with the following contents:

#!/bin/sh
[ -p "$FIFO_UEBERZUG" ] && printf '{"action":"remove","identifier":"preview"}\n' >"$FIFO_UEBERZUG"

Next, make a file named previewer in the lf configuration directory. Your previewer script may look like this:

#!/bin/sh
draw() {
  path="$(readlink -f -- "$1" | sed 's/\\/\\\\/g;s/"/\\"/g')"
  printf '{"action":"add","identifier":"preview","x":%d,"y":%d,"width":%d,"height":%d,"scaler":"contain","scaling_position_x":0.5,"scaling_position_y":0.5,"path":"%s"}\n' \
    "$x" "$y" "$width" "$height" "$path" >"$FIFO_UEBERZUG"
  exit 1
}

hash() {
  cache="$HOME/.cache/lf/$(stat --printf '%n\0%i\0%F\0%s\0%W\0%Y' -- "$(readlink -f -- "$1")" | sha256sum | cut -d' ' -f1).jpg"
}

cache() {
  if ! [ -f "$cache" ]; then
    dir="$(dirname -- "$cache")"
    [ -d "$dir" ] || mkdir -p -- "$dir"
    "$@"
  fi
  draw "$cache"
}

file="$1"
width="$2"
height="$3"
x="$4"
y="$5"

case "$(file -Lb --mime-type -- "$file")" in
  image/*)
    if [ -p "$FIFO_UEBERZUG" ]; then
      # ueberzug doesn't handle image orientation correctly
      orientation="$(magick identify -format '%[orientation]\n' -- "$file[0]")"
      if [ -n "$orientation" ] \
          && [ "$orientation" != Undefined ] \
          && [ "$orientation" != TopLeft ]; then
        hash "$file"
        cache magick -- "$file[0]" -auto-orient "$cache"
      else
        draw "$file"
      fi
    fi
    ;;
  video/*)
    if [ -p "$FIFO_UEBERZUG" ]; then
      hash "$file"
      cache ffmpegthumbnailer -i "$file" -o "$cache" -s 0
    fi
    ;;
  text/*)
    exec cat "$file"
    ;;
esac

file -Lb -- "$file" | fold -s -w "$width"
exit 0

Using a preview cache in ~/.cache/lf is entirely optional. A path under /tmp/ may be used instead.

Make sure all of the above files are executable. Take note that the draw() function exits with code 1. This is to signal lf not to cache the result of the previewer script so that the next time the user selects the same file the previewer script will be executed again.

Finally, place the following lines in your lfrc:

set previewer ~/.config/lf/previewer
set cleaner ~/.config/lf/cleaner

More extensive examples are available at the following repositories:

With Kitty and Pistol

The following setup will use kitty to display images, and fall back to pistol for everything else. Other terminals like wezterm and Konsole also implement Kitty's image protocol, so those can be substituted if desired.

As usual, we'll specify a previewer and a cleaner in ~/.config/lf/lfrc:

set previewer ~/.config/lf/previewer
set cleaner ~/.config/lf/cleaner

The cleaner script is ~/.config/lf/cleaner:

#!/bin/sh
exec kitty +kitten icat --clear --stdin no --silent --transfer-mode file </dev/null >/dev/tty

And the previewer is ~/.config/lf/previewer:

#!/bin/sh
draw() {
  kitty +kitten icat --silent --stdin no --transfer-mode file --place "${w}x${h}@${x}x${y}" "$1" </dev/null >/dev/tty
  exit 1
}

file="$1"
w="$2"
h="$3"
x="$4"
y="$5"

case "$(file -Lb --mime-type "$file")" in 
  image/*)
    draw "$file"
    ;;
  video/*)
    # vidthumb is from here:
    # https://raw.githubusercontent.com/duganchen/kitty-pistol-previewer/main/vidthumb
    draw "$(vidthumb "$file")"
    ;;
esac

pistol "$file"

Note that this example uses the vidthumb script.

With Sixel

If your terminal has Sixel support, you can display image previews in Sixel format:

  1. Install chafa or any similar program that can convert images to Sixel.
  2. Set the sixel and previewer options in your config file.
  3. Add the following previewer script:
#!/bin/sh
case "$(file -Lb --mime-type -- "$1")" in
  image/*)
    chafa -f sixel -s "$2x$3" --animate off --polite on "$1"
    exit 1
    ;;
  text/*)
    cat "$1"
    ;;
esac

Note

This only works partially for KDE konsole (images can't be cleaned up automatically, only by redrawing the screen with ctrl+l). In the official Sixel manual, printing new characters over an image is undefined, but most other terminals will overwrite the image.

Note

Sixel previews will appear distorted if chafa is run in a new session (e.g. bwrap sandbox with the --new-session option). This is because chafa is unable to obtain the terminal cell size in pixels, which is necessary to calculate the pixel size of the Sixel image.

Note

tmux does not display Sixel images over 1MB by default but you can change the limit with INPUT_BUF_LIMIT in input.c.

With stpv and ctpv

stpv and ctpv are previewer utilities made for integrating into lf. No wrapper scripts are needed. You only need to add 4 lines in lf config to make either one of them work.

stpv is a POSIX shell script, while ctpv is a rewrite of stpv written in C. ctpv is faster and has a few additional features.

stpv:

set previewer stpv
set cleaner stpvimgclr
&stpvimg --listen $id
cmd on-quit $stpvimg --end $id

ctpv:

set previewer ctpv
set cleaner ctpvclear
&ctpv -s $id
&ctpvquit $id

With w3mimgdisplay

Note

Previews using w3mimgdisplay are not persistent (scrolling, making selections, or moving the window will erase the image), since they are drawn directly on top of the terminal window. Unless you are a framebuffer user, it is not recommended to use w3mimgdisplay.

Make a file named draw_img.sh in the lf configuration folder with the contents below:

#!/bin/bash

clear_screen() {
    printf '\e[%sH\e[9999C\e[1J%b\e[1;%sr' \
           "$((LINES-2))" "${TMUX:+\e[2J}" "$max_items"
}

# Get a file's mime_type.
mime_type=$(file -bi "$1")

# File isn't an image file, give warning.
if [[ $mime_type != image/* ]]; then
    lf -remote "send $id echoerr 'Not an image'"
    exit
fi

w3m_paths=(/usr/{local/,}{lib,libexec,lib64,libexec64}/w3m/w3mi*)
read -r w3m _ < <(type -p w3mimgdisplay "${w3m_paths[@]}")
read -r LINES COLUMNS < <(stty size)

# Get terminal window size in pixels and set it to WIDTH and HEIGHT.
export $(xdotool getactivewindow getwindowgeometry --shell)

# Get the image size in pixels.
read -r img_width img_height < <("$w3m" <<< "5;${CACHE:-$1}")

((img_width > WIDTH)) && {
    ((img_height=img_height*WIDTH/img_width))
    ((img_width=WIDTH))
}

((img_height > HEIGHT)) && {
    ((img_width=img_width*HEIGHT/img_height))
    ((img_height=HEIGHT))
}

# Variable needed for centering image.
HALF_HEIGHT=$(expr $HEIGHT / 2)
HALF_WIDTH=$(expr $WIDTH / 2)
HALF_IMG_HEIGHT=$(expr $img_height / 2)
HALF_IMG_WIDTH=$(expr $img_width / 2)
X_POS=$(expr $HALF_WIDTH - $HALF_IMG_WIDTH)
Y_POS=$(expr $HALF_HEIGHT - $HALF_IMG_HEIGHT)

clear_screen
# Hide the cursor.
printf '\e[?25l'

# Display the image.
printf '0;1;%s;%s;%s;%s;;;;;%s\n3;\n4\n' \
    ${X_POS:-0} \
    ${Y_POS:-0} \
    "$img_width" \
    "$img_height" \
    "${CACHE:-$1}" | "$w3m" &>/dev/null

# Wait for user input.
read -ern 1

# Clear the image.
printf '6;%s;%s;%s;%s\n3;' \
    "${X_POS:-0}" \
    "${Y_POS:-0}" \
    "$WIDTH" \
    "$HEIGHT" | "$w3m" &>/dev/null

clear_screen

Now add key mappings corresponding to the script.

map - $~/.config/lf/draw_img.sh "$f"

And likewise for video previews:

cmd video_preview ${{
    cache="$(mktemp "${TMPDIR:-/tmp}/thumb_cache.XXXXX")"
    ffmpegthumbnailer -i "$f" -o "$cache" -s 0
    ~/.config/lf/draw_img.sh "$cache"
}}
map + :video_preview

Sandboxing preview operations

While the default lf configuration only previews text files, using more complex preview parsers is somewhat dangerous. In case there is a vulnerability in a preview parser like pdftotext, it is possible to use this simple script to sandbox the previewer:

#!/bin/bash
## ~/.config/lf/previewer_sandbox

set -euo pipefail
(
    exec bwrap \
     --ro-bind /usr/bin /usr/bin \
     --ro-bind /usr/share/ /usr/share/ \
     --ro-bind /usr/lib /usr/lib \
     --ro-bind /usr/lib64 /usr/lib64 \
     --symlink /usr/bin /bin \
     --symlink /usr/bin /sbin \
     --symlink /usr/lib /lib \
     --symlink /usr/lib64 /lib64 \
     --proc /proc \
     --dev /dev  \
     --ro-bind /etc /etc \
     --ro-bind ~/.config ~/.config \
     --ro-bind ~/.cache ~/.cache \
     --ro-bind "$PWD" "$PWD" \
     --unshare-all \
     --new-session \
     bash ~/.config/lf/preview.sh "$@"
)

Set your previewer to the sandbox script and have your real preview script at ~/.config/lf/previewer

set previewer ~/.config/lf/previewer_sandbox
Clone this wiki locally