extract transcript from YouTube video

yt-transcript.sh

#!/bin/bash
# yt-transcript.sh - Extract transcripts from YouTube videos
# 
# Author: Marc Hanheide

set -e

# Default values
LANG="en"
FORMAT="plain"
OUTPUT_FILE=""
KEEP_TEMP=0
TEMP_DIR="/tmp"

# Function to display usage information
usage() {
    cat << EOF
Usage: $(basename "$0") [OPTIONS] VIDEO_URL

Extract transcript from a YouTube video.

Options:
  -h, --help              Show this help message and exit
  -l, --language LANG     Language code for transcript (default: en)
  -f, --format FORMAT     Output format: plain, srt, or json (default: plain)
  -o, --output FILE       Output file (default: print to stdout)
  -k, --keep-temp         Keep temporary files
  -t, --temp-dir DIR      Temporary directory (default: /tmp)

Example:
  $(basename "$0") https://www.youtube.com/watch?v=dQw4w9WgXcQ
  $(basename "$0") -l fr -f srt -o transcript.srt https://www.youtube.com/watch?v=dQw4w9WgXcQ
EOF
    exit 1
}

# Function to cleanup temporary files
cleanup() {
    if [ $KEEP_TEMP -eq 0 ]; then
        rm -f "${TEMP_DIR}/transcript".*
    fi
}

# Parse command line arguments
POSITIONAL=()
while [[ $# -gt 0 ]]; do
    key="$1"
    case $key in
        -h|--help)
            usage
            ;;
        -l|--language)
            LANG="$2"
            shift 2
            ;;
        -f|--format)
            FORMAT="$2"
            shift 2
            ;;
        -o|--output)
            OUTPUT_FILE="$2"
            shift 2
            ;;
        -k|--keep-temp)
            KEEP_TEMP=1
            shift
            ;;
        -t|--temp-dir)
            TEMP_DIR="$2"
            shift 2
            ;;
        -*)
            echo "Error: Unknown option $1"
            usage
            ;;
        *)
            POSITIONAL+=("$1")
            shift
            ;;
    esac
done

# Check if a video URL was provided
if [ ${#POSITIONAL[@]} -eq 0 ]; then
    echo "Error: No YouTube URL provided"
    usage
fi

# Set the video URL
VIDEO_URL=${POSITIONAL[0]}
TEMP_BASE="${TEMP_DIR}/transcript"

# Register cleanup on exit
trap cleanup EXIT

yt-dlp --get-description --skip-download "$VIDEO_URL" > "${TEMP_BASE}.description"

# Download subtitles
yt-dlp --skip-download --write-subs --write-auto-subs \
    --sub-lang "$LANG" --sub-format ttml --convert-subs srt \
    --output "${TEMP_BASE}.%(ext)s" "$VIDEO_URL"  1>&2

# Process according to requested format
case $FORMAT in
    srt)
        # Just use the SRT file as is
        RESULT_CONTENT=$(cat "${TEMP_BASE}.${LANG}.srt")
        ;;
    json)
        # Convert to a basic JSON format
        RESULT_CONTENT=$(
            # Read description into variable with proper escaping
            DESCRIPTION=$(cat "${TEMP_BASE}.description" | sed 's/"/\\"/g' | tr '\n' ' ')
            
            echo "{"
            echo "  \"description\": \"${DESCRIPTION}\","
            echo "  \"transcript\": ["
            awk 'BEGIN{RS=""; FS="\n"} {
                gsub(/^[0-9]+$/, "", $1);
                gsub(/<[^>]*>/, "", $3);
                gsub(/\"/, "\\\"", $3);
                if (NR > 1) printf ",\n";
                printf "    {\"time\": \"%s\", \"text\": \"%s\"}", $2, $3;
            }' "${TEMP_BASE}.${LANG}.srt"
            echo -e "\n  ]"
            echo "}"
        )
        ;;
    plain|*)
        # Extract plain text
        RESULT_CONTENT=$(
            # First get the description
            cat "${TEMP_BASE}.description"
            echo -e "\n\n"
            # Then get the transcript
            cat "${TEMP_BASE}.${LANG}.srt" | 
            sed '/^$/d' | 
            grep -v '^[0-9]*$' | 
            grep -v '\-->' | 
            sed 's/<[^>]*>//g' | 
            tr '\n' ' '
        )
        ;;
esac

# Output result
if [ -n "$OUTPUT_FILE" ]; then
    echo "$RESULT_CONTENT" > "$OUTPUT_FILE"
    echo "Transcript saved to $OUTPUT_FILE" >&2
else
    echo "$RESULT_CONTENT"
fi

requires https://github.com/yt-dlp/yt-dlp to be installed