Integrate transcript captions into captions API

config/config.example.yml

@@ -182,6 +182,19 @@ https_only: false
 #force_resolve:
 
 
+##
+## Use Innertube's transcripts API instead of timedtext for closed captions
+##
+## Useful for larger instances as InnerTube is **not ratelimited**. See https://github.com/iv-org/invidious/issues/2567
+##
+## Subtitle experience may differ slightly on Invidious.  
+##
+## Accepted values: true, false
+## Default: false
+##
+# use_innertube_for_captions: false
+
+
 # -----------------------------
 #  Logging
 # -----------------------------

src/invidious/config.cr

@@ -129,6 +129,9 @@ class Config
   # Use quic transport for youtube api
   property use_quic : Bool = false
 
+  # Use Innertube's transcripts API instead of timedtext for closed captions
+  property use_innertube_for_captions : Bool = false
+
   # Saved cookies in "name1=value1; name2=value2..." format
   @[YAML::Field(converter: Preferences::StringToCookies)]
   property cookies : HTTP::Cookies = HTTP::Cookies.new

src/invidious/routes/api/v1/videos.cr

@@ -87,6 +87,13 @@ module Invidious::Routes::API::V1::Videos
       caption = caption[0]
     end
 
+    if CONFIG.use_innertube_for_captions
+      params = Invidious::Videos::Transcript.generate_param(id, caption.language_code, caption.auto_generated)
+      initial_data = YoutubeAPI.transcript(params.to_s)
+
+      webvtt = Invidious::Videos::Transcript.convert_transcripts_to_vtt(initial_data, caption.language_code)
+    else
+      # Timedtext API handling
       url = URI.parse("#{caption.base_url}&tlang=#{tlang}").request_target
 
       # Auto-generated captions often have cues that aren't aligned properly with the video,
@@ -153,6 +160,7 @@ module Invidious::Routes::API::V1::Videos
             .gsub(/([0-9:.]{12} --> [0-9:.]{12}).+/, "\\1")
         end
       end
+    end
 
     if title = env.params.query["title"]?
       # https://blog.fastmail.com/2011/06/24/download-non-english-filenames/

src/invidious/videos/caption.cr

@@ -6,7 +6,9 @@ module Invidious::Videos
     property language_code : String
     property base_url : String
 
-    def initialize(@name, @language_code, @base_url)
+    property auto_generated : Bool
+
+    def initialize(@name, @language_code, @base_url, @auto_generated)
     end
 
     # Parse the JSON structure from Youtube
@@ -25,7 +27,12 @@ module Invidious::Videos
         language_code = caption["languageCode"].to_s
         base_url = caption["baseUrl"].to_s
 
-        captions_list << CaptionMetadata.new(name, language_code, base_url)
+        auto_generated = false
+        if caption["kind"]? && caption["kind"] == "asr"
+          auto_generated = true
+        end
+
+        captions_list << CaptionMetadata.new(name, language_code, base_url, auto_generated)
       end
 
       return captions_list

src/invidious/videos/transcript.cr

@@ -85,7 +85,13 @@ module Invidious::Videos
 
       lines = [] of TranscriptLine
       body.each do |line|
+        # Transcript section headers. They are not apart of the captions and as such we can safely skip them.
+        if line.as_h.has_key?("transcriptSectionHeaderRenderer")
+          next
+        end
+
         line = line["transcriptSegmentRenderer"]
+
         start_ms = line["startMs"].as_s.to_i.millisecond
         end_ms = line["endMs"].as_s.to_i.millisecond