Ruby script for Windows and Linux to transcode or copy essential media tracks into a smaller, more portable format while remaining high enough quality to be mistaken for the original.

special-transcode.rb

#!/usr/bin/env ruby
#
# special-transcode.rb
#
# Copyright (c) 2019-2022 Don Melton
#

require 'English'
require 'fileutils'
require 'json'
require 'optparse'

module Transcoding

  class UsageError < RuntimeError
  end

  class Command
    def about
      <<-HERE
special-transcode.rb 0.0.02022101201
Copyright (c) 2019-2022 Don Melton
      HERE
    end

    def usage
      <<-HERE
Transcode or copy essential media tracks into a smaller, more portable format
while remaining high enough quality to be mistaken for the original.

Usage: #{File.basename($PROGRAM_NAME)} [OPTION]... [FILE]...

Creates a Matroska `.mkv` format file in the current working directory
with video in 10-bit HEVC format and audio in multichannel AAC format.
Forced subtitles are automatically included along with one other subtitle,
the SDH version being preferred.

Options:
    --position TIME, --duration TIME
                      start transcoding at position and/or limit to duration
                        in seconds[.milliseconds] or [HH:]MM:SS[.m...] format
    --debug           increase diagnostic information and create `.log` file
-n, --dry-run         don't transcode, just show `ffmpeg` command and exit
-p, --preview         create incomplete output viewable while transcoding
-q, --cq VALUE        set constant quality value (default: 27)
-b, --bitrate TARGET  use average bitrate instead of constant quality
    --preset NAME     apply video encoder preset
    --no-cuda         disable CUDA video decoder and pipeline
    --no-bframe-refs  don't set mode for using B-frames as reference frames
    --deinterlace     reduce interlace artifacts without changing frame rate
                        (applied automatically for some inputs)
    --blur            reduce noise with temporary change in resolution
    --fdk-vbr MODE    set FDK AAC audio encoder variable bitrate (VBR) mode
                        (default: 5, use 0 to disable VBR)
    --eac3            use Dolby Digital Plus (E-AC-3) instead of AAC audio
    --language CODE   match subtitles using language code in ISO 639-2 format
                        (default: eng)
    --name STRING     match first subtitle track with name containing string
                        when SDH or track without name can't be found
                        (default: english)
-h, --help            display this help and exit
    --version         output version information and exit

Requires `ffprobe`, `ffmpeg`, `mkvextract`, `mkvmerge` and `mkvpropedit`.
      HERE
    end

    def initialize
      @position = nil
      @duration = nil
      @debug = false
      @dry_run = false
      @preview = false
      @cq = '27'
      @bitrate = nil
      @preset = nil
      @cuda = true
      @bframe_refs = true
      @deinterlace = false
      @blur = false
      @aac_encoder = 'aac'
      @fdk_vbr_mode = '5'
      @eac3 = false
      @language = 'eng'
      @name = 'english'
    end

    def run
      begin
        OptionParser.new do |opts|
          define_options opts

          opts.on '-h', '--help' do
            puts usage
            exit
          end

          opts.on '--version' do
            puts about
            exit
          end
        end.parse!
      rescue OptionParser::ParseError => e
        raise UsageError, e
      end

      fail UsageError, 'missing argument' if ARGV.empty?

      configure
      ARGV.each { |arg| process_input arg }
      exit
    rescue UsageError => e
      Kernel.warn "#{$PROGRAM_NAME}: #{e}"
      Kernel.warn "Try `#{File.basename($PROGRAM_NAME)} --help` for more information."
      exit false
    rescue StandardError => e
      Kernel.warn "#{$PROGRAM_NAME}: #{e}"
      exit(-1)
    rescue SignalException
      puts
      exit(-1)
    end

    def define_options(opts)
      opts.on '--position ARG' do |arg|
        @position = resolve_time(arg)
      end

      opts.on '--duration ARG' do |arg|
        @duration = resolve_time(arg)
      end

      opts.on '--debug' do
        @debug = true
      end

      opts.on '-n', '--dry-run' do
        @dry_run = true
      end

      opts.on '-p', '--preview' do
        @preview = true
      end

      opts.on '-q', '--cq ARG', Float do |arg|
        @cq = [[arg, 0].max, 51].min.to_s.sub(/\.0$/, '')
        @bitrate = nil
      end

      opts.on '-b', '--bitrate ARG', Integer do |arg|
        @bitrate = arg
      end

      opts.on '--preset ARG' do |arg|
        @preset = case arg
        when 'slow', 'medium', 'fast', /p[1-7]/
          arg
        else
          fail UsageError, "invalid preset: #{arg}"
        end
      end

      opts.on '--no-cuda' do
        @cuda = false
      end

      opts.on '--no-bframe-refs' do
        @bframe_refs = false
      end

      opts.on '--deinterlace' do
        @deinterlace = true
      end

      opts.on '--blur' do
        @blur = true
      end

      opts.on '--fdk-vbr ARG', Integer do |arg|
        @fdk_vbr_mode = [[arg, 0].max, 5].min.to_s
        @eac3 = false
      end

      opts.on '--eac3' do
        @eac3 = true
      end

      opts.on '--language ARG' do |arg|
        fail UsageError, "invalid subtitle language code: #{arg}" unless arg =~ /^[a-z]{3}$/

        @language = arg
      end

      opts.on '--name ARG' do |arg|
        @name = arg
      end
    end

    def resolve_time(arg)
      time = 0.0

      case arg
      when /^([0-9]+(?:\.[0-9]+)?)$/
        time = $1.to_f
      when /^(?:(?:([0-9][0-9]):)?([0-9][0-9]):)?([0-9][0-9](?:\.[0-9]+)?)$/
        time = $3.to_f
        time = ($2.to_i * 60) + time unless $2.nil?
        time = ($1.to_i * 60 * 60) + time unless $1.nil?
      else
        fail UsageError, "invalid time: #{arg}"
      end

      time
    end

    def configure
      Kernel.warn 'Configuring...'
      encoders = ''

      IO.popen([
        'ffmpeg',
        '-loglevel', 'quiet',
        '-encoders'
      ]) do |io|
        encoders = io.read
      end

      fail 'configuring failed' unless $CHILD_STATUS.exitstatus == 0

      @aac_encoder = 'libfdk_aac' if encoders =~ /libfdk_aac/
    end

    def process_input(path)
      seconds = Time.now.tv_sec
      media_info = scan_media(path)
      video = nil
      first_audio = nil
      audio = nil
      forced_subtitle = nil
      sdh_subtitle = nil
      other_subtitle = nil
      named_subtitle = nil

      media_info['streams'].each do |stream|
        case stream['codec_type']
        when 'video'
          video = stream if video.nil?
        when 'audio'
          first_audio = stream if first_audio.nil?
          audio = stream if audio.nil? and stream['disposition']['default'] == 1
        when 'subtitle'
          next if stream.fetch('tags', {}).fetch('language', '') != @language

          if stream['disposition']['forced'] == 1
            forced_subtitle = stream if forced_subtitle.nil?
          else
            title = stream.fetch('tags', {}).fetch('title', '')
            sdh_subtitle = stream if sdh_subtitle.nil? and title =~ /sdh/i
            other_subtitle = stream if other_subtitle.nil? and title.empty?
            named_subtitle = stream if named_subtitle.nil? and title =~ /#{@name}/i
          end
        end
      end

      fail "video track not found: #{arg}" if video.nil?

      audio ||= first_audio
      fail "audio track not found: #{arg}" if audio.nil?

      other_subtitle ||= named_subtitle
      time_options = get_time_options(media_info)
      pix_fmt = video.fetch('pix_fmt', 'yuv420p')

      if @cuda and pix_fmt == 'yuv420p'
        decoding_options = [
          '-hwaccel', 'cuda',
          '-hwaccel_output_format', 'cuda'
        ]
      else
        decoding_options = video['codec_name'] == 'vc1' ? ['-hwaccel', 'auto'] : []
      end

      Kernel.warn 'Stream mapping:'
      video_options = get_video_options(video)
      audio_options = get_audio_options(audio)
      subtitle_options = get_subtitle_options(forced_subtitle, sdh_subtitle, other_subtitle)
      output = File.basename(path, '.*') + ".#{@preview ? '_PREVIEW_.mkv' : 'mp4'}"
      fail "output file already exists: #{output}" if File.exist? output

      ffmpeg_command = [
        'ffmpeg',
        '-loglevel', (@debug ? 'verbose' : 'error'),
        '-stats',
        *time_options,
        *decoding_options,
        '-i', path,
        *video_options,
        *audio_options,
        '-sn',
        *(@preview ? [] : ['-map_chapters', '-1']),
        '-metadata:g', 'title=',
        *(@preview ? ['-default_mode', 'passthrough'] : []),
        output
      ]

      if @preview or subtitle_options.empty?
        subtitles = nil
      else
        subtitles = File.basename(path) + '.subtitles.mkv'
        fail "subtitles file already exists: #{subtitles}" if File.exist? subtitles

        ffmpeg_command += [
          '-vn',
          '-an',
          *subtitle_options,
          '-map_chapters', '-1',
          subtitles
        ]
      end

      command_line = escape_command(ffmpeg_command)
      Kernel.warn 'Command line:'

      if @dry_run
        puts command_line
        return
      end

      Kernel.warn command_line
      mkv_output = File.basename(path, '.*') + '.mkv'

      unless @preview
        fail "output file already exists: #{mkv_output}" if File.exist? mkv_output
      end

      if media_info['chapters'].empty?
        chapters = nil
      else
        chapters = File.basename(path) + '.chapters.xml'

        unless @preview
          fail "chapters file already exists: #{chapters}" if File.exist? chapters
        end
      end

      Kernel.warn 'Transcoding...'
      fail "transcoding failed: #{output}" unless system(
        (@debug ? {'FFREPORT' => 'level=40'} : {}),
        *ffmpeg_command
      )

      return if @preview

      Kernel.warn 'Multiplexing...'
      hdr_options = []

      if pix_fmt != 'yuv420p'
        hdr_info = ''

        IO.popen([
          'ffprobe',
          '-loglevel', 'quiet',
          '-select_streams', 'v:0',
          '-show_frames',
          '-read_intervals', '%+#1',
          '-show_entries', 'frame=side_data_list',
          '-print_format', 'json',
          path
        ]) do |io|
          hdr_info = io.read
        end

        fail "scanning media failed: #{path}" unless $CHILD_STATUS.exitstatus == 0

        begin
          hdr_info = JSON.parse(hdr_info)
        rescue JSON::JSONError
          fail "HDR information not found: #{path}"
        end

        md = nil
        cll = nil

        hdr_info['frames'].each do |frame|
          frame.fetch('side_data_list', []).each do |side_data|
            if side_data['side_data_type'] == 'Mastering display metadata'
              md = side_data if md.nil?
            elsif side_data['side_data_type'] == 'Content light level metadata'
              cll = side_data if cll.nil?
            end
          end
        end

        unless md.nil? or cll.nil?
          hdr_options = [
            '--max-content-light',        "0:#{cll['max_content']}",
            '--max-frame-light',          "0:#{cll['max_average']}",
            '--chromaticity-coordinates', "0:#{eval md['red_x'] + '.0'}," +
                                            "#{eval md['red_y'] + '.0'}," +
                                            "#{eval md['green_x'] + '.0'}," +
                                            "#{eval md['green_y'] + '.0'}," +
                                            "#{eval md['blue_x'] + '.0'}," +
                                            "#{eval md['blue_y'] + '.0'}",
            '--white-colour-coordinates', "0:#{eval md['white_point_x'] + '.0'}," +
                                            "#{eval md['white_point_y'] + '.0'}",
            '--max-luminance',            "0:#{eval md['max_luminance'] + '.0'}",
            '--min-luminance',            "0:#{eval md['min_luminance'] + '.0'}"
          ]
        end
      end

      if chapters.nil?
        adjustment = nil
      else
        fail "chapters file already exists: #{chapters}" if File.exist? chapters

        content = ''

        IO.popen(['mkvextract', path, 'chapters']) do |io|
          index = 0

          io.each do |line|
            if line =~ /<ChapterString>/
              index += 1
              line.sub!(/(#{$MATCH})[^<]*/, '\1' + "Chapter #{index}")
            end

            content += line
          end
        end

        fail "chapter extraction failed: #{path}" unless $CHILD_STATUS.exitstatus == 0

        begin
          chapters_file = File.new(chapters, 'wb')
          chapters_file.print content
          chapters_file.close
        rescue SystemCallError => e
          raise "writing chapters file failed: #{e}"
        end

        if @position.nil?
          adjustment = nil
        else
          adjustment = (([media_info['format']['duration'].to_f - 1.0, @position].min) * -1000).to_i.to_s
        end
      end

      fail "output file already exists: #{mkv_output}" if File.exist? mkv_output

      fail "multiplexing failed: #{mkv_output}" unless system(
        'mkvmerge',
        '--output', mkv_output,
        *hdr_options,
        output,
        *(subtitles.nil? ? [] : ['--no-track-tags', '--no-global-tags', subtitles]),
        *(adjustment.nil? ? [] : ['--chapter-sync', adjustment]),
        *(chapters.nil? ? [] : ['--chapters', chapters])
      )

      FileUtils.rm output
      FileUtils.rm subtitles unless subtitles.nil?
      FileUtils.rm chapters unless chapters.nil?

      fail "property editing failed: #{mkv_output}" unless system(
        'mkvpropedit',
        '--quiet',
        mkv_output,
        '--set', 'muxing-application=',
        '--set', 'writing-application='
      )

      Kernel.warn "\nElapsed time: #{seconds_to_time(Time.now.tv_sec - seconds)}\n\n"
    end

    def scan_media(path)
      Kernel.warn 'Scanning media...'
      media_info = ''

      IO.popen([
        'ffprobe',
        '-loglevel', 'quiet',
        '-show_streams',
        '-show_chapters',
        '-show_format',
        '-print_format', 'json',
        path
      ]) do |io|
        media_info = io.read
      end

      fail "scanning media failed: #{path}" unless $CHILD_STATUS.exitstatus == 0

      begin
        media_info = JSON.parse(media_info)
      rescue JSON::JSONError
        fail "media information not found: #{path}"
      end

      Kernel.warn media_info.inspect if @debug
      media_info
    end

    def escape_command(command)
      command_line = ''
      command.each {|item| command_line += "#{escape_string(item)} " }
      command_line.sub!(/ $/, '')
      command_line
    end

    def escape_string(str)
      # See: https://github.com/larskanis/shellwords
      return '""' if str.empty?

      str = str.dup

      if RUBY_PLATFORM =~ /mingw/
        str.gsub!(/((?:\\)*)"/) { "\\" * ($1.length * 2) + "\\\"" }

        if str =~ /\s/
          str.gsub!(/(\\+)\z/) { "\\" * ($1.length * 2 ) }
          str = "\"#{str}\""
        end
      else
        str.gsub!(/([^A-Za-z0-9_\-.,:\/@\n])/, "\\\\\\1")
        str.gsub!(/\n/, "'\n'")
      end

      str
    end

    def get_time_options(media_info)
      duration = media_info['format']['duration'].to_f
      fail "media duration too short: #{duration}" if duration < 2.0

      if @position.nil?
        position = 0.0
      else
        position = [duration - 1.0, @position].min
        duration -= position
      end

      duration = [duration, [@duration, 0.1].max].min unless @duration.nil?
      options = []
      options += ['-ss', position.to_s.sub(/\.0$/, '')] unless @position.nil?
      options += ['-t', duration.to_s.sub(/\.0$/, '')] unless @duration.nil?
      time = seconds_to_time(duration.to_i)
      milliseconds = duration.to_s.sub(/^[0-9]+(\.[0-9]+)$/, '\1')
      time += milliseconds unless milliseconds == '.0'
      Kernel.warn "duration = #{time}"
      options
    end

    def seconds_to_time(seconds)
      sprintf("%02d:%02d:%02d", seconds / (60 * 60), (seconds / 60) % 60, seconds % 60)
    end

    def get_video_options(video)
      if video['codec_name'] == 'mpeg2video' and video['avg_frame_rate'] == '30000/1001'
        options = ['-vsync', 'cfr']
      else
        options = []
      end

      pix_fmt = video.fetch('pix_fmt', 'yuv420p')

      if @cuda and pix_fmt == 'yuv420p'
        cuda = true
        scale_filter = 'scale_cuda'
        yadif_filter = 'yadif_cuda'
      else
        cuda = false
        scale_filter = 'scale'
        yadif_filter = 'yadif'
      end

      width   = video['width'].to_i
      height  = video['height'].to_i

      if @blur
        filters = ["#{scale_filter}=#{(width / 6) * 5}:#{(height / 6) * 5}", "#{scale_filter}=#{width}:#{height}"]
      else
        filters = []
      end

      filters += ['scale_cuda=format=p010le'] if cuda
      filters += [yadif_filter] if @deinterlace or video.fetch('field_order', 'progressive') != 'progressive'

      if filters.empty?
        options += ['-map', "0:#{video['index']}"]
      else
        options += [
          '-filter_complex', "[0:#{video['index']}]#{filters.join(',')}[v]",
          '-map', '[v]',
        ]
      end

      color_primaries = video['color_primaries']
      color_trc       = video['color_transfer']
      colorspace      = video['color_space']

      if pix_fmt == 'yuv420p10le'
        color_primaries ||= 'bt2020'
        color_trc       ||= 'smpte2084'
        colorspace      ||= 'bt2020nc'
      end

      if width > 1920 or height > 1080
        maxrate = 25000
      elsif width > 1280 or height > 720
        maxrate = 12000
      elsif width > 720 or height > 576
        maxrate = 10000
      else
        color_primaries ||= (width == 720 and height == 576 and video['codec_name'] == 'mpeg2video') ? 'bt470bg' : 'smpte170m'
        colorspace      ||= 'smpte170m'
        maxrate = 6000
      end

      color_primaries ||= 'bt709'
      color_trc       ||= 'bt709'
      colorspace      ||= 'bt709'
      bitrate = [[@bitrate, 100].max, maxrate - 100].min unless @bitrate.nil?
      Kernel.warn "#{sprintf("%2d", video['index'])} = " + (@bitrate.nil? ? "#{@cq} CQ" : "#{bitrate} Kbps") + ' video'

      options += [
        *(cuda ? [] : ['-pix_fmt:v', 'p010le']),
        '-c:v', 'hevc_nvenc',
        *(@bitrate.nil? ? ['-cq:v', @cq] : ['-b:v', "#{bitrate}k"]),
        '-maxrate:v', "#{maxrate}k",
        '-bufsize:v', "#{maxrate}k",
        *(@preset.nil? ? [] : ['-preset:v', @preset]),
        '-spatial-aq:v', '1',
        '-rc-lookahead:v', '32',
        *(@bframe_refs ? ['-b_ref_mode:v', 'middle'] : []),
        *(@preview ? [] : ['-bsf:v', 'filter_units=remove_types=35|38-40']),
        '-color_primaries:v', color_primaries,
        '-color_trc:v', color_trc,
        '-colorspace:v', colorspace,
        '-metadata:s:v', 'title=',
        '-disposition:v', 'default'
      ]
    end

    def get_audio_options(audio)
      codec = audio['codec_name']
      channels = audio['channels'].to_i

      if @eac3
        encoder = codec =~ /ac3/ ? 'copy' : 'eac3'
      else
        encoder = (codec == 'aac' and channels <= 6) ? 'copy' : @aac_encoder
      end

      Kernel.warn "#{sprintf("%2d", audio['index'])} = #{encoder} audio"

      options = [
        '-map', "0:#{audio['index']}",
        '-c:a', encoder,
        *((encoder == 'libfdk_aac' and @fdk_vbr_mode != '0') ? ['-vbr:a', @fdk_vbr_mode] : []),
        *((encoder != 'copy' and channels > 2) ? (@eac3 ? ['-b:a', '640k'] : ['-ac:a', '6']) : []),
        *((encoder != 'copy' and channels > 6 and @eac3) ? ['-ac:a', '6'] : []),
        *((encoder != 'copy' and audio['sample_rate'] != '48000') ? ['-ar:a', '48000'] : []),
        '-metadata:s:a', 'title=',
        '-disposition:a', 'default'
      ]
    end

    def get_subtitle_options(forced_subtitle, sdh_subtitle, other_subtitle)
      options = []
      index = 0

      unless forced_subtitle.nil?
        Kernel.warn "#{sprintf("%2d", forced_subtitle['index'])} = Forced subtitle"

        options += [
          '-map', "0:#{forced_subtitle['index']}",
          "-c:s:#{index}", 'copy',
          "-metadata:s:s:#{index}", 'title=Forced',
          "-disposition:s:#{index}", 'default+forced'
        ]

        index += 1
      end

      if sdh_subtitle.nil?
        unless other_subtitle.nil?
          Kernel.warn "#{sprintf("%2d", other_subtitle['index'])} = (no name) subtitle"

          options += [
            '-map', "0:#{other_subtitle['index']}",
            "-c:s:#{index}", 'copy',
            "-metadata:s:s:#{index}", 'title=',
            "-disposition:s:#{index}", '0'
          ]
        end
      else
        Kernel.warn "#{sprintf("%2d", sdh_subtitle['index'])} = SDH subtitle"

        options += [
          '-map', "0:#{sdh_subtitle['index']}",
          "-c:s:#{index}", 'copy',
          "-metadata:s:s:#{index}", 'title=SDH',
          "-disposition:s:#{index}", '0'
        ]
      end

      options
    end
  end
end

Transcoding::Command.new.run