audio.rb

file = open(ARGV[0])

require 'tempfile'
require 'fileutils'

ary = []

file.each_line do |line|
  text = ''
  line.split(' ').each do |word|
    if word =~ /[\p{Han}\p{Hiragana}\p{Katakana}，．、。ー・]+/
      if text != ''
        ary << {lang: 'en', text: text.chop}
      end

      text = ''
      ary << {lang: 'jp', text: word}
    else
      text += word + ' '
    end
  end

  if text != ''
    ary << {lang: 'en', text: text.chop}
  end
end

ary.each_with_index do |hash,i|
  wav = "#{i}.wav"
  puts "#{hash[:lang]}: #{hash[:text]} @#{wav} #{i.to_i+1}/#{ary.count}"
  case hash[:lang]
  when 'en'
    file = Tempfile.create 'base'
    file.puts hash[:text]
    file.close
    system 'text2wave', file.path, '-o', wav
  when 'jp'
    file = Tempfile.create 'base'
    file.puts hash[:text].sub(/^　/, '')
    file.close
    system 'open_jtalk', '-ow', wav, '-x', '/var/lib/mecab/dic/open-jtalk/naist-jdic', '-m', '/usr/share/hts-voice/nitech-jp-atr503-m001/nitech_jp_atr503_m001.htsvoice', '-s', '48000', '-r', '1.5', file.path
  end
end

(0..ary.count-1).each{|i|
  wav = "#{i}.wav"
  if File.exists? wav
    system 'sox', wav, '-r', '48000', 'shhhhhhhhhhhhhhhhit.wav'
    FileUtils.rm wav
    FileUtils.mv 'shhhhhhhhhhhhhhhhit.wav', wav rescue nil
  end
  puts "#{i.to_i+1}/#{ary.count}"
}
system "sox #{(0..ary.count-1).map{|i| "#{i}.wav" }.select{|x| File.exists? x}.join(" ")} '#{ARGV[1]}'"
(0..ary.count-1).each{|i| FileUtils.rm "#{i}.wav" rescue nil}