merge_to_opus.rb

merge_to_opus.rb

#!/usr/bin/env ruby
require 'digest'

#
# A script that merges multiple audio files to a single OPUS File with chapter
# markers based on the input files.
# Every input file will be one chapter in the output OPUS File.
# The ouptut file is hardcoded to be 'merged.opus'
#
# For more information about OPUS Chapter marks:
# => https://wiki.xiph.org/Chapter_Extension
#
# For more information about the ffmpeg concat filter:
# => https://trac.ffmpeg.org/wiki/Concatenate
# => https://ffmpeg.org/ffmpeg-filters.html#concat
#
# USAGE: ./merge_to_opus.rb file1.flac file2.flac file3.flac ...
# EXAMPLE: ./merge_to_opus Chapter_*.flac
# AUTHOR: Johannes Hertenstein, 2017
# LICENSE: WTFPL
#

#
# General settings.
# The output file format and options can be set here.
#
options = [
	'-acodec libopus',
	'-b:a 64000',
	'-vbr on',
	'-compression_level 10',
	'-safe 0'
]

#
# Formats a time in seconds according to the OPUS chapter marks specification.
# The output will always be a string of fixed length with the following structure:
# HH:MM:SS.SSS
#
def format_time(seconds)
	minutes = 0
	hours = 0

	while seconds >= 60
		minutes += 1
		seconds -= 60
	end

	while minutes >= 60
		hours += 1
		minutes -= 60
	end

	hours = hours.to_s.rjust(2,'0')
	minutes = minutes.to_s.rjust(2,'0')
	seconds = sprintf('%.3f', seconds).rjust(6, '0')

	return "#{hours}:#{minutes}:#{seconds}"
end

total_length=0
chapter=1
files = ARGV

files.each do |file|
	# Find out the time of the current chapter
	seconds = `ffprobe -i #{file} -show_entries format=duration -v quiet -of csv="p=0"`.to_i
	time = format_time(total_length)
	total_length += seconds

	# Extract a chapter name from the filename
	name = file.split('.')[0...-1].join('.')

	# format the chapter number
	num = chapter.to_s.rjust(3, '0')

	# Add the metadata options to ffmpegs optionlist
	options.push("-metadata CHAPTER#{num}=#{time}")
	options.push("-metadata CHAPTER#{num}NAME=#{name}")

	chapter += 1
end

# build filter_complex map. This map specifies which stream of which input
# file should be mapped to which stream of the output. For us this is easy:
# The input files have one audio stream as does the output.
filter_complex_map = (0...files.length).map{ |i| "[#{i}:a:0]" }.join(' ')
filter_complex = filter_complex_map + ' concat=n=' + files.length.to_s + ':v=0:a=1 [a]'
input_files = files.map{ |f| "-i #{f}" }.join(" \\\n\t ")

# Finally: Execute ffmpeg
cmd = "ffmpeg #{input_files} " +
				" -filter_complex \"#{filter_complex}\" \\\n\t " +
				" -map '[a]' \\\n\t " +
				options.join(" \\\n\t ") +
				" \\\n\t merged.opus "
puts "$ #{cmd}"
system(cmd)

This version does not support filenames with spaces. I added support for such filenames by enclosing some of the generated arguments in (escaped) parentheses :) See the diff below. Maybe merge this with your version? :)

67c67
< 	seconds = `ffprobe -i #{file} -show_entries format=duration -v quiet -of csv="p=0"`.to_i
---
> 	seconds = `ffprobe -i "#{file}" -show_entries format=duration -v quiet -of csv="p=0"`.to_i
78,79c78,79
< 	options.push("-metadata CHAPTER#{num}=#{time}")
< 	options.push("-metadata CHAPTER#{num}NAME=#{name}")
---
> 	options.push("-metadata \"CHAPTER#{num}=#{time}\"")
> 	options.push("-metadata \"CHAPTER#{num}NAME=#{name}\"")
89c89
< input_files = files.map{ |f| "-i #{f}" }.join(" \\\n\t ")
---
> input_files = files.map{ |f| "-i \"#{f}\"" }.join(" \\\n\t ")