txt-to-mp3: Convert txt to mp3 with Amazon Polly. txt-to-mp3 （テキストファイルを分割してAmazon Pollyでmp3に変換し、ffmpegを使って合成するRubyスクリプト）

config.yaml

MP3DIR: ./mp3
WHO: Mizuki

input.txt

<!-- opening.mp3 -->

おはようございます。

<!-- chime1.mp3 -->

こんにちは。

<!-- chime2.mp3 -->

おやすみなさい。

<!-- ending.mp3 -->

txt-to-mp3

#!/usr/bin/env ruby

require 'fileutils'
require 'tmpdir'
require 'pp'
require 'yaml'

APPNAME = 'txt-to-mp3'

if ARGV.length != 3
  puts "Usage: txt-to-mp3 config.yaml input.txt output.mp3"
  puts
  puts "input.txt の例"
  puts <<-"EOD"
■input.txt の例: 

<!-- opening.mp3 -->

おはようございます。

<!-- chime1.mp3 -->

こんにちは。

<!-- chime2.mp3 -->

おやすみなさい。

<!-- ending.mp3 -->

■config.yaml の例: 
MP3DIR: ./mp3
WHO: Mizuki
  EOD
  abort
end

CONFIG = YAML.load_file(ARGV[0])

def extract_mp3(line)
  if line.match(/^<!-- (.+\.mp3) -->/)
    $1
  else
    nil
  end
end

def split_txt_to_sections(input_txt)
  mp3dir = CONFIG['MP3DIR']
  if not Dir.exist?(mp3dir)
    abort "#{APPNAME}: #{mp3dir} is not found."
  end
  sections = []
  lines = IO.readlines(input_txt)
  cursection = ''
  lines.each do |line|
    mp3 = extract_mp3(line)
    if mp3
      sections << { :type => :text, :content => cursection }
      cursection = ''
      sections << { :type => :mp3, :content => "#{mp3dir}/#{mp3}" }
    else
      cursection += line
    end
  end
  if cursection.size > 0
    sections << { :type => :text, :content => cursection }
  end
  sections
end

def sections_to_files(sections, tmpdir)
  index = 1
  files = []
  sections.each do |section|
    if section[:type] == :text
      filename = sprintf("#{tmpdir}/_%d.txt", index)
      open(filename, "w") do |f|
        f.print section[:content]
      end
      files << { :type => :text, :filename => filename }
      index += 1
    elsif section[:type] == :mp3
      files << { :type => :mp3, :filename => section[:content] }
    else
      abort "#{APPNAME}: sections_to_files: invalid type: " + section[:type].to_s
    end
  end
  files
end

def convert_files_to_mp3files(files)
  mp3files = []
  files.each do |file|
    if file[:type] == :text
      filename = file[:filename]
      mp3filename = filename + '.mp3'
      puts "Converting: " + filename
      cmd = "aws polly synthesize-speech --output-format mp3 --text file://#{filename} --voice-id #{CONFIG['WHO']} #{mp3filename}"
      puts cmd
      err = system(cmd)
      puts err
      mp3files << mp3filename
    elsif file[:type] == :mp3
      mp3files << file[:filename]
    else
      abort "#{APPNAME}: convert_txt_to_mp3: invalid type: " + file[:type].to_s
    end
  end
  mp3files
end

def main(input_txt, output_mp3)
  puts "Input text file: #{input_txt}"
  puts "Output mp3 file: #{output_mp3}"
  puts "MP3 dir: #{CONFIG['MP3DIR']}"
  Dir.mktmpdir do |tmpdir|
    sections = split_txt_to_sections(input_txt)
    files = sections_to_files(sections, tmpdir)
    mp3files = convert_files_to_mp3files(files)
    mp3s = mp3files.join('|')
    cmd = "ffmpeg -y -i \"concat:#{mp3s}\" #{output_mp3}"
    puts cmd
    result = system(cmd)
    puts result
  end
end

main(ARGV[1], ARGV[2])