mp3から要約テキストを得るまでの一連の作業をまとめたシェルスクリプト

summerize-mp3-book.sh

#!/bin/bash -v
BOOK_NAME=$1

# 1ファイルに結合する
ls *.mp3 \
| sort -u \
| awk '{print "file " $0 "" }' \
> ${BOOK_NAME}.mp3-list.txt
ffmpeg \
    -f concat \
    -safe 0 \
    -i ${BOOK_NAME}.mp3-list.txt \
    -c copy ${BOOK_NAME}.mp3

# 時間で分割する
ffmpeg -i ${BOOK_NAME}.mp3 \
    -f segment -segment_time 900 -c copy \
${BOOK_NAME}.%03d.mp3

# 文字起こし
ls ${BOOK_NAME}.*.mp3 \
| sort -u \
| while read MP3_FILE_NAME; do
    # 文字起こし
    echo "generating speech ${MP3_FILE_NAME} to text ${MP3_FILE_NAME}.transcribe.txt"
    curl \
        --request POST \
        --silent \
        --url https://api.openai.com/v1/audio/transcriptions \
        --header "Authorization: Bearer ${OPENAI_API_KEY}" \
        --header "Content-Type: multipart/form-data" \
        --form model=whisper-1 \
        --form file=@${MP3_FILE_NAME} \
    > ${MP3_FILE_NAME}.transcribe.txt

    # API Rate limit を考慮して待つ
    sleep 60;
done

# 要約
ls ${BOOK_NAME}.*.mp3 \
| sort -u \
| while read MP3_FILE_NAME; do
    # 要約
    echo "summerize text ${MP3_FILE_NAME}.transcribe.txt to ${MP3_FILE_NAME}.summerize.txt"
    cat ${MP3_FILE_NAME}.transcribe.txt \
    | python3 ~/bin/summerize-stdin-by-bedrock.py \
    > ${MP3_FILE_NAME}.summerize.txt

    # API Rate limit を考慮して待つ
    sleep 60;
done

# 全要約テキストをさらに要約する
cat ${BOOK_NAME}.*.summerize.txt \
| python3 ~/bin/summerize-stdin-by-bedrock.py \
> ${BOOK_NAME}.summerize.txt