pdf-ocr

#!/usr/bin/env bash
# Prerequisites (all can be brew installed):
#   pdf2image qpdf tesseract

replace=false

usage() {
    cat << EOF
usage: $0 [-r] file1 ...

Convert each (pdf) file to one that has the hidden (searchable) text layer.

If the -r option is specified, each file will be REPLACED.
If the -r option is NOT specified, a new PDF will be created.
Example:
- original file: sample.pdf
- new file: sample.ocr.pdf
EOF
    exit 1
}

if [ $# -eq 0 ]; then
    usage
fi

convert() {
    inFile="$1"
    if $replace; then
        outFile="$inFile"
    else
        # shellcheck disable=SC2001
        outFile=$(echo "$inFile" | sed 's/^\(.*\)\(\.[^.]*\)$/\1.ocr\2/')
    fi
    echo "Converting file $inFile to $outFile"

    if ! [ -f "$inFile" ]; then
        >&2 echo "error: specified input pdf file '$inFile' not found"
        exit 1
    fi

    tmpDir=$(mktemp -d -t pdf)
    cp "$inFile" "$tmpDir/$inFile"
    pushd "$tmpDir" || exit 1

    pdf2image -zoom 3 "$inFile" JUNK
    for p in JUNK*.png; do tesseract "$p" "$p" -l eng pdf; done
    popd || exit 1
    qpdf --empty --pages "$tmpDir"/JUNK*.pdf -- "$outFile"
    rm -rf "$tmpDir"
}


while [ $# -gt 0 ]; do
    if [ "$1" = "-r" ]; then
        replace=true
    else
        convert "$1"
    fi
    shift
done