diff --git a/src/invidious/frontend/watch_page.cr b/src/invidious/frontend/watch_page.cr
index b860dba7..5fd81168 100644
--- a/src/invidious/frontend/watch_page.cr
+++ b/src/invidious/frontend/watch_page.cr
@@ -7,7 +7,7 @@ module Invidious::Frontend::WatchPage
     getter full_videos : Array(Hash(String, JSON::Any))
     getter video_streams : Array(Hash(String, JSON::Any))
     getter audio_streams : Array(Hash(String, JSON::Any))
-    getter captions : Array(Invidious::Videos::CaptionMetadata)
+    getter captions : Array(Invidious::Videos::Captions::Metadata)
 
     def initialize(
       @full_videos,
diff --git a/src/invidious/videos.cr b/src/invidious/videos.cr
index 2b1d2603..9fbd1374 100644
--- a/src/invidious/videos.cr
+++ b/src/invidious/videos.cr
@@ -24,7 +24,7 @@ struct Video
   property updated : Time
 
   @[DB::Field(ignore: true)]
-  @captions = [] of Invidious::Videos::CaptionMetadata
+  @captions = [] of Invidious::Videos::Captions::Metadata
 
   @[DB::Field(ignore: true)]
   property adaptive_fmts : Array(Hash(String, JSON::Any))?
@@ -215,9 +215,9 @@ struct Video
     keywords.includes? "YouTube Red"
   end
 
-  def captions : Array(Invidious::Videos::CaptionMetadata)
+  def captions : Array(Invidious::Videos::Captions::Metadata)
     if @captions.empty? && @info.has_key?("captions")
-      @captions = Invidious::Videos::CaptionMetadata.from_yt_json(info["captions"])
+      @captions = Invidious::Videos::Captions::Metadata.from_yt_json(info["captions"])
     end
 
     return @captions
diff --git a/src/invidious/videos/caption.cr b/src/invidious/videos/caption.cr
index 1e2abde9..82b68dcd 100644
--- a/src/invidious/videos/caption.cr
+++ b/src/invidious/videos/caption.cr
@@ -1,107 +1,109 @@
 require "json"
 
 module Invidious::Videos
-  struct CaptionMetadata
-    property name : String
-    property language_code : String
-    property base_url : String
+  module Captions
+    struct Metadata
+      property name : String
+      property language_code : String
+      property base_url : String
 
-    property auto_generated : Bool
+      property auto_generated : Bool
 
-    def initialize(@name, @language_code, @base_url, @auto_generated)
-    end
-
-    # Parse the JSON structure from Youtube
-    def self.from_yt_json(container : JSON::Any) : Array(CaptionMetadata)
-      caption_tracks = container
-        .dig?("playerCaptionsTracklistRenderer", "captionTracks")
-        .try &.as_a
-
-      captions_list = [] of CaptionMetadata
-      return captions_list if caption_tracks.nil?
-
-      caption_tracks.each do |caption|
-        name = caption["name"]["simpleText"]? || caption["name"]["runs"][0]["text"]
-        name = name.to_s.split(" - ")[0]
-
-        language_code = caption["languageCode"].to_s
-        base_url = caption["baseUrl"].to_s
-
-        auto_generated = false
-        if caption["kind"]? && caption["kind"] == "asr"
-          auto_generated = true
-        end
-
-        captions_list << CaptionMetadata.new(name, language_code, base_url, auto_generated)
+      def initialize(@name, @language_code, @base_url, @auto_generated)
       end
 
-      return captions_list
-    end
+      # Parse the JSON structure from Youtube
+      def self.from_yt_json(container : JSON::Any) : Array(Captions::Metadata)
+        caption_tracks = container
+          .dig?("playerCaptionsTracklistRenderer", "captionTracks")
+          .try &.as_a
 
-    def timedtext_to_vtt(timedtext : String, tlang = nil) : String
-      # In the future, we could just directly work with the url. This is more of a POC
-      cues = [] of XML::Node
-      tree = XML.parse(timedtext)
-      tree = tree.children.first
+        captions_list = [] of Captions::Metadata
+        return captions_list if caption_tracks.nil?
 
-      tree.children.each do |item|
-        if item.name == "body"
-          item.children.each do |cue|
-            if cue.name == "p" && !(cue.children.size == 1 && cue.children[0].content == "\n")
-              cues << cue
+        caption_tracks.each do |caption|
+          name = caption["name"]["simpleText"]? || caption["name"]["runs"][0]["text"]
+          name = name.to_s.split(" - ")[0]
+
+          language_code = caption["languageCode"].to_s
+          base_url = caption["baseUrl"].to_s
+
+          auto_generated = false
+          if caption["kind"]? && caption["kind"] == "asr"
+            auto_generated = true
+          end
+
+          captions_list << Captions::Metadata.new(name, language_code, base_url, auto_generated)
+        end
+
+        return captions_list
+      end
+
+      def timedtext_to_vtt(timedtext : String, tlang = nil) : String
+        # In the future, we could just directly work with the url. This is more of a POC
+        cues = [] of XML::Node
+        tree = XML.parse(timedtext)
+        tree = tree.children.first
+
+        tree.children.each do |item|
+          if item.name == "body"
+            item.children.each do |cue|
+              if cue.name == "p" && !(cue.children.size == 1 && cue.children[0].content == "\n")
+                cues << cue
+              end
             end
+            break
           end
-          break
         end
-      end
-      result = String.build do |result|
-        result << <<-END_VTT
-        WEBVTT
-        Kind: captions
-        Language: #{tlang || @language_code}
+        result = String.build do |result|
+          result << <<-END_VTT
+          WEBVTT
+          Kind: captions
+          Language: #{tlang || @language_code}
 
 
-        END_VTT
+          END_VTT
 
-        result << "\n\n"
+          result << "\n\n"
 
-        cues.each_with_index do |node, i|
-          start_time = node["t"].to_f.milliseconds
+          cues.each_with_index do |node, i|
+            start_time = node["t"].to_f.milliseconds
 
-          duration = node["d"]?.try &.to_f.milliseconds
+            duration = node["d"]?.try &.to_f.milliseconds
 
-          duration ||= start_time
+            duration ||= start_time
 
-          if cues.size > i + 1
-            end_time = cues[i + 1]["t"].to_f.milliseconds
-          else
-            end_time = start_time + duration
+            if cues.size > i + 1
+              end_time = cues[i + 1]["t"].to_f.milliseconds
+            else
+              end_time = start_time + duration
+            end
+
+            # start_time
+            result << start_time.hours.to_s.rjust(2, '0')
+            result << ':' << start_time.minutes.to_s.rjust(2, '0')
+            result << ':' << start_time.seconds.to_s.rjust(2, '0')
+            result << '.' << start_time.milliseconds.to_s.rjust(3, '0')
+
+            result << " --> "
+
+            # end_time
+            result << end_time.hours.to_s.rjust(2, '0')
+            result << ':' << end_time.minutes.to_s.rjust(2, '0')
+            result << ':' << end_time.seconds.to_s.rjust(2, '0')
+            result << '.' << end_time.milliseconds.to_s.rjust(3, '0')
+
+            result << "\n"
+
+            node.children.each do |s|
+              result << s.content
+            end
+            result << "\n"
+            result << "\n"
           end
-
-          # start_time
-          result << start_time.hours.to_s.rjust(2, '0')
-          result << ':' << start_time.minutes.to_s.rjust(2, '0')
-          result << ':' << start_time.seconds.to_s.rjust(2, '0')
-          result << '.' << start_time.milliseconds.to_s.rjust(3, '0')
-
-          result << " --> "
-
-          # end_time
-          result << end_time.hours.to_s.rjust(2, '0')
-          result << ':' << end_time.minutes.to_s.rjust(2, '0')
-          result << ':' << end_time.seconds.to_s.rjust(2, '0')
-          result << '.' << end_time.milliseconds.to_s.rjust(3, '0')
-
-          result << "\n"
-
-          node.children.each do |s|
-            result << s.content
-          end
-          result << "\n"
-          result << "\n"
         end
+        return result
       end
-      return result
     end
 
     # List of all caption languages available on Youtube.
diff --git a/src/invidious/videos/transcript.cr b/src/invidious/videos/transcript.cr
index ba2728cd..c86b3988 100644
--- a/src/invidious/videos/transcript.cr
+++ b/src/invidious/videos/transcript.cr
@@ -37,7 +37,7 @@ module Invidious::Videos
       # Convert into array of TranscriptLine
       lines = self.parse(initial_data)
 
-      # Taken from Invidious::Videos::CaptionMetadata.timedtext_to_vtt()
+      # Taken from Invidious::Videos::Captions::Metadata.timedtext_to_vtt()
       vtt = String.build do |vtt|
         vtt << <<-END_VTT
         WEBVTT
diff --git a/src/invidious/views/user/preferences.ecr b/src/invidious/views/user/preferences.ecr
index b1061ee8..55349c5a 100644
--- a/src/invidious/views/user/preferences.ecr
+++ b/src/invidious/views/user/preferences.ecr
@@ -89,7 +89,7 @@
                 <label for="captions[0]"><%= translate(locale, "preferences_captions_label") %></label>
                 <% preferences.captions.each_with_index do |caption, index| %>
                     <select class="pure-u-1-6" name="captions[<%= index %>]" id="captions[<%= index %>]">
-                        <% Invidious::Videos::CaptionMetadata::LANGUAGES.each do |option| %>
+                        <% Invidious::Videos::Captions::LANGUAGES.each do |option| %>
                             <option value="<%= option %>" <% if preferences.captions[index] == option %> selected <% end %>><%= translate(locale, option.blank? ? "none" : option) %></option>
                         <% end %>
                     </select>