Extract embedded captions while processing file

lib/active_encode/engine_adapters/ffmpeg_adapter.rb

@@ -42,6 +42,7 @@ def create(input_url, options = {})
         # Create a working directory that holds all output files related to the encode
         FileUtils.mkdir_p working_path("", new_encode.id)
         FileUtils.mkdir_p working_path("outputs", new_encode.id)
+        FileUtils.mkdir_p working_path("supplemental_files", new_encode.id)
 
         # Extract technical metadata from input file
         curl_option = if options && options[:headers]
@@ -91,6 +92,8 @@ def create(input_url, options = {})
           new_encode.input.duration = fixed_duration(working_path("duration_input_metadata", new_encode.id))
         end
 
+        options[:subtitle_count] = new_encode.input.subtitle_count if new_encode.input.subtitle_count&.positive?
+
         new_encode.state = :running
         new_encode.percent_complete = 1
         new_encode.errors = []
@@ -150,6 +153,7 @@ def find(id, opts = {})
         end
 
         encode.output = build_outputs encode if encode.completed?
+        encode.output += build_supplemental_outputs encode if encode.completed?
 
         encode
       end
@@ -251,17 +255,46 @@ def build_outputs(encode)
         outputs
       end
 
+      def build_supplemental_outputs(encode)
+        id = encode.id
+        files = []
+        Dir["#{File.absolute_path(working_path('supplemental_files', id))}/*"].each do |file_path|
+          file = ActiveEncode::Output.new
+          file.url = "file://#{file_path}"
+          file.id = "#{encode.input.id}-#{File.basename(file_path)}"
+          file.created_at = encode.created_at
+          file.updated_at = File.mtime file_path
+
+          files << file
+        end
+
+        files
+      end
+
       def ffmpeg_command(input_url, id, opts)
+        sanitized_filename = ActiveEncode.sanitize_base input_url
         output_opt = opts[:outputs].collect do |output|
-          sanitized_filename = ActiveEncode.sanitize_base input_url
           file_name = "outputs/#{sanitized_filename}-#{output[:label]}.#{output[:extension]}"
           " #{output[:ffmpeg_opt]} #{working_path(file_name, id)}"
         end.join(" ")
+
+        supplemental_file_opt = caption_extraction_options(sanitized_filename, opts[:subtitle_count], id) if opts[:subtitle_count]&.positive?
+
         header_opt = Array(opts[:headers]).map do |k, v|
           "#{k}: #{v}\r\n"
         end.join
         header_opt = "-headers '#{header_opt}'" if header_opt.present?
-        "#{FFMPEG_PATH} #{header_opt} -y -loglevel level+fatal -progress #{working_path('progress', id)} -i \"#{input_url}\" #{output_opt}"
+        "#{FFMPEG_PATH} #{header_opt} -y -loglevel level+fatal -progress #{working_path('progress', id)} -i \"#{input_url}\" #{supplemental_file_opt} #{output_opt}"
+      end
+
+      def caption_extraction_options(filename, count, id)
+        opts = ""
+        (0..count - 1).each do |i|
+          subtitle_filename = "supplemental_files/#{filename}-caption#{i}.vtt"
+          opts += " -map 0:s:#{i} -c:s webvtt #{working_path(subtitle_filename, id)}"
+        end
+
+        opts
       end
 
       def get_pid(id)
@@ -314,6 +347,7 @@ def get_tech_metadata(file_path)
         doc.remove_namespaces!
         duration = get_xpath_text(doc, '//Duration/text()', :to_f)
         duration *= 1000 unless duration.nil? # Convert to milliseconds
+        subtitle_count = doc.xpath('//track[@type="Text"]').length
         { url: get_xpath_text(doc, '//media/@ref', :to_s),
           width: get_xpath_text(doc, '//Width/text()', :to_f),
           height: get_xpath_text(doc, '//Height/text()', :to_f),
@@ -323,7 +357,8 @@ def get_tech_metadata(file_path)
           audio_codec: get_xpath_text(doc, '//track[@type="Audio"]/CodecID/text()', :to_s),
           audio_bitrate: get_xpath_text(doc, '//track[@type="Audio"]/BitRate/text()', :to_i),
           video_codec: get_xpath_text(doc, '//track[@type="Video"]/CodecID/text()', :to_s),
-          video_bitrate: get_xpath_text(doc, '//track[@type="Video"]/BitRate/text()', :to_i) }
+          video_bitrate: get_xpath_text(doc, '//track[@type="Video"]/BitRate/text()', :to_i),
+          subtitle_count: subtitle_count }
       end
 
       def get_xpath_text(doc, xpath, cast_method)

lib/active_encode/engine_adapters/ffmpeg_adapter/cleaner.rb

@@ -53,7 +53,11 @@ def remove_files(files, older_than)
         def remove_empty_directories(directories)
           directories_to_delete = directories.select { |d| Dir.empty?(d) }
           non_empty_directories = directories - directories_to_delete
-          directories_to_delete += non_empty_directories.select { |ned| Dir.children(ned) == ["outputs"] && directories_to_delete.include?(File.join(ned, "outputs")) }
+          directories_to_delete += non_empty_directories.select do |ned|
+            Dir.children(ned).sort == ["outputs", "supplemental_files"] &&
+              directories_to_delete.include?(File.join(ned, "outputs")) &&
+              directories_to_delete.include?(File.join(ned, "supplemental_files"))
+          end
           FileUtils.rmdir(directories_to_delete) unless directories_to_delete.empty?
         end
 

lib/active_encode/technical_metadata.rb

@@ -22,11 +22,13 @@ module TechnicalMetadata
       attr_accessor :video_codec
       attr_accessor :audio_bitrate
       attr_accessor :video_bitrate
+
+      attr_accessor :subtitle_count
     end
 
     def assign_tech_metadata(metadata)
       [:width, :height, :frame_rate, :duration, :file_size, :checksum,
-       :audio_codec, :video_codec, :audio_bitrate, :video_bitrate].each do |field|
+       :audio_codec, :video_codec, :audio_bitrate, :video_bitrate, :subtitle_count].each do |field|
         send("#{field}=", metadata[field]) if metadata.key?(field)
       end
     end

spec/integration/ffmpeg_adapter_spec.rb

@@ -322,6 +322,42 @@ def find_encode(id)
       end
     end
 
+    context 'file with embedded captions' do
+      let(:file_with_embedded_captions) { "file://" + Rails.root.join('..', 'spec', 'fixtures', 'file_with_embedded_captions.mp4').to_s }
+      let!(:create_embedded_captions_job) { ActiveEncode::Base.create(file_with_embedded_captions, outputs: [{ label: "low", ffmpeg_opt: "-s 640x480", extension: 'mp4' }]) }
+      let(:find_embedded_captions_job) { ActiveEncode::Base.find create_embedded_captions_job.id }
+
+      it "does not have errors" do
+        sleep 2
+        expect(find_embedded_captions_job.errors).to be_empty
+      end
+
+      it "has the input technical metadata in a file" do
+        expect(File.read("#{work_dir}/#{create_embedded_captions_job.id}/input_metadata")).not_to be_empty
+      end
+
+      it "has the pid in a file" do
+        expect(File.read("#{work_dir}/#{create_embedded_captions_job.id}/pid")).not_to be_empty
+      end
+
+      context 'when uri encoded' do
+        let(:file_with_embedded_captions) { Addressable::URI.encode("file://" + Rails.root.join('..', 'spec', 'fixtures', 'file_with_embedded_captions.mp4').to_s) }
+
+        it "does not have errors" do
+          sleep 2
+          expect(find_embedded_captions_job.errors).to be_empty
+        end
+
+        it "has the input technical metadata in a file" do
+          expect(File.read("#{work_dir}/#{create_embedded_captions_job.id}/input_metadata")).not_to be_empty
+        end
+
+        it "has the pid in a file" do
+          expect(File.read("#{work_dir}/#{create_embedded_captions_job.id}/pid")).not_to be_empty
+        end
+      end
+    end
+
     context 'when failed' do
       subject { created_job }
 
@@ -505,6 +541,10 @@ def find_encode(id)
     end
 
     context ":all" do
+      # The cleaner removes empty directories even if they are younger than the defined :older_than param.
+      # We need to use a file that will populate the supplemental files directory for proper testing of :all behavior.
+      let(:file) { "file://" + Rails.root.join('..', 'spec', 'fixtures', 'file_with_embedded_captions.mp4').to_s }
+
       it "deletes all files and directories older than 2 weeks" do
         sleep 1
         travel 3.weeks do