[frostschutz]

Here's how I create flashcard images. (this is very specific to the data I exported from my Anki kanji deck, but the idea in general would work for other things, too)

First you write an SVG file and style it up the way you want:

<svg width="1080" height="1440">
  <style>
    .kanji { font-family: "EPSON 教科書体Ｍ";
             font-size: 900px;
    }
    .reading1 { font-family: "EPSON 教科書体Ｍ";
               font-size: 50px;
    }
    .reading2 { font-family: "EPSON 教科書体Ｍ";
               font-size: 50px;
    }
    .meaning1 { font-family: "Linux Libertine Display O";
                font-size: 50px;
                font-style: italic;
    }
    .meaning2 { font-family: "Linux Libertine Display O";
                font-size: 50px;
    }
    .graphem { font-family: "Linux Libertine Display O";
               font-size: 40px;
    }
  </style>
  <text class="kanji" x="50%" y="54%" alignment-baseline="middle" text-anchor="middle">KANJI</text>
  <text class="reading1" x="50%" y="66%" alignment-baseline="middle" text-anchor="middle">READING1</text>
  <text class="reading2" x="50%" y="71%" alignment-baseline="middle" text-anchor="middle">READING2</text>
  <text class="meaning1" x="50%" y="80%" alignment-baseline="middle" text-anchor="middle">MEANING1</text>
  <text class="meaning2" x="50%" y="85%" alignment-baseline="middle" text-anchor="middle">MEANING2</text>
  <text class="graphem" x="50%" y="95%" alignment-baseline="middle" text-anchor="middle">GRAPHEM</text>
</svg>

Then you shell script the heck out of it:

#!/bin/bash

while read line
do
    kanji=$(printf "%s" "$line" | cut -f1)
    meaning=$(printf "%s" "$line" | cut -f2)
    reading=$(printf "%s" "$line" | cut -f3)
    graphem=$(printf "%s" "$line" | cut -f4)

    reading1=${reading%'<br>'*}
    reading2=""

    if [ "$reading1" != "$reading" ]
    then
        reading2=${reading#*'<br>'}
    fi

    meaning1=${meaning%'<br>'*}

    if [ "$meaning1" != "$meaning" ]
    then
        meaning2=${meaning#*'<br>'}
    else
        meaning1=""
        meaning2=$meaning
    fi

    echo $kanji , $reading1, $reading2 , $meaning1, $meaning2 , $graphem

    # generate question cards
    sed -r -e "s@KANJI@${kanji}@g" \
           -e 's@(READING[12]|MEANING[12]|GRAPHEM)@@g' \
           kanji.svg > question.svg
    inkscape --export-png=question/"$kanji".png question.svg > /dev/null &

    # generate answer cards
    sed -r -e "s@KANJI@@g" \
           -e "s@READING1@${reading1}@g" \
           -e "s@READING2@${reading2}@g" \
           -e "s@MEANING1@${meaning1}@g" \
           -e "s@MEANING2@${meaning2}@g" \
           -e "s@GRAPHEM@${graphem}@g" \
           kanji.svg > answer.svg
    inkscape --export-png=answer/"$kanji".png answer.svg > /dev/null &

    wait
    rm question.svg answer.svg

    # optimize
    ( convert -flatten question/"$kanji".png +dither -colors 2 png8:question/"$kanji".convert.png
      optipng -quiet question/"$kanji".convert.png
      mv question/"$kanji".convert.png question/"$kanji".png
    ) &

    ( convert answer/"$kanji".png +dither -colors 2 png8:answer/"$kanji".convert.png
      optipng -quiet answer/"$kanji".convert.png
      mv answer/"$kanji".convert.png answer/"$kanji".png
    ) &

    # wait # optimize can go into the next loop
done < kanji.txt

wait

And that's about it.