2022-05-24 05:37:58 +09:00
|
|
|
require "json"
|
|
|
|
|
|
|
|
module Invidious::Videos
|
2023-08-25 08:00:02 +09:00
|
|
|
module Captions
|
|
|
|
struct Metadata
|
|
|
|
property name : String
|
|
|
|
property language_code : String
|
|
|
|
property base_url : String
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
property auto_generated : Bool
|
2023-07-23 21:02:02 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
def initialize(@name, @language_code, @base_url, @auto_generated)
|
|
|
|
end
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
# 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
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
captions_list = [] of Captions::Metadata
|
|
|
|
return captions_list if caption_tracks.nil?
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
caption_tracks.each do |caption|
|
|
|
|
name = caption["name"]["simpleText"]? || caption["name"]["runs"][0]["text"]
|
|
|
|
name = name.to_s.split(" - ")[0]
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
language_code = caption["languageCode"].to_s
|
|
|
|
base_url = caption["baseUrl"].to_s
|
2022-05-24 05:37:58 +09:00
|
|
|
|
2023-08-25 08:21:05 +09:00
|
|
|
auto_generated = (caption["kind"]? == "asr")
|
2023-08-25 08:00:02 +09:00
|
|
|
|
|
|
|
captions_list << Captions::Metadata.new(name, language_code, base_url, auto_generated)
|
2023-07-23 21:02:02 +09:00
|
|
|
end
|
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
return captions_list
|
2022-05-24 05:37:58 +09:00
|
|
|
end
|
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
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
|
2023-01-04 00:21:16 +09:00
|
|
|
end
|
2023-08-25 08:00:02 +09:00
|
|
|
break
|
2023-01-04 00:21:16 +09:00
|
|
|
end
|
|
|
|
end
|
2023-01-09 06:18:35 +09:00
|
|
|
|
2023-08-25 07:10:50 +09:00
|
|
|
settings_field = {
|
|
|
|
"Kind" => "captions",
|
|
|
|
"Language" => "#{tlang || @language_code}",
|
|
|
|
}
|
2023-01-09 06:18:35 +09:00
|
|
|
|
2023-08-25 07:10:50 +09:00
|
|
|
result = WebVTT.build(settings_field) do |vtt|
|
2023-08-25 08:00:02 +09:00
|
|
|
cues.each_with_index do |node, i|
|
|
|
|
start_time = node["t"].to_f.milliseconds
|
2023-01-04 00:21:16 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
duration = node["d"]?.try &.to_f.milliseconds
|
2023-01-04 00:21:16 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
duration ||= start_time
|
2023-01-04 00:21:16 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
if cues.size > i + 1
|
|
|
|
end_time = cues[i + 1]["t"].to_f.milliseconds
|
|
|
|
else
|
|
|
|
end_time = start_time + duration
|
|
|
|
end
|
2023-01-04 00:21:16 +09:00
|
|
|
|
2023-08-25 07:10:50 +09:00
|
|
|
text = String.build do |io|
|
|
|
|
node.children.each do |s|
|
|
|
|
io << s.content
|
|
|
|
end
|
2023-08-25 08:00:02 +09:00
|
|
|
end
|
2023-08-25 07:10:50 +09:00
|
|
|
|
2023-08-25 07:42:42 +09:00
|
|
|
vtt.cue(start_time, end_time, text)
|
2023-01-04 03:10:26 +09:00
|
|
|
end
|
2023-01-04 00:21:16 +09:00
|
|
|
end
|
2023-08-25 07:10:50 +09:00
|
|
|
|
2023-08-25 08:00:02 +09:00
|
|
|
return result
|
2023-01-04 00:21:16 +09:00
|
|
|
end
|
|
|
|
end
|
|
|
|
|
2022-05-24 05:37:58 +09:00
|
|
|
# List of all caption languages available on Youtube.
|
|
|
|
LANGUAGES = {
|
|
|
|
"",
|
|
|
|
"English",
|
|
|
|
"English (auto-generated)",
|
|
|
|
"English (United Kingdom)",
|
|
|
|
"English (United States)",
|
|
|
|
"Afrikaans",
|
|
|
|
"Albanian",
|
|
|
|
"Amharic",
|
|
|
|
"Arabic",
|
|
|
|
"Armenian",
|
|
|
|
"Azerbaijani",
|
|
|
|
"Bangla",
|
|
|
|
"Basque",
|
|
|
|
"Belarusian",
|
|
|
|
"Bosnian",
|
|
|
|
"Bulgarian",
|
|
|
|
"Burmese",
|
|
|
|
"Cantonese (Hong Kong)",
|
|
|
|
"Catalan",
|
|
|
|
"Cebuano",
|
|
|
|
"Chinese",
|
|
|
|
"Chinese (China)",
|
|
|
|
"Chinese (Hong Kong)",
|
|
|
|
"Chinese (Simplified)",
|
|
|
|
"Chinese (Taiwan)",
|
|
|
|
"Chinese (Traditional)",
|
|
|
|
"Corsican",
|
|
|
|
"Croatian",
|
|
|
|
"Czech",
|
|
|
|
"Danish",
|
|
|
|
"Dutch",
|
|
|
|
"Dutch (auto-generated)",
|
|
|
|
"Esperanto",
|
|
|
|
"Estonian",
|
|
|
|
"Filipino",
|
|
|
|
"Finnish",
|
|
|
|
"French",
|
|
|
|
"French (auto-generated)",
|
|
|
|
"Galician",
|
|
|
|
"Georgian",
|
|
|
|
"German",
|
|
|
|
"German (auto-generated)",
|
|
|
|
"Greek",
|
|
|
|
"Gujarati",
|
|
|
|
"Haitian Creole",
|
|
|
|
"Hausa",
|
|
|
|
"Hawaiian",
|
|
|
|
"Hebrew",
|
|
|
|
"Hindi",
|
|
|
|
"Hmong",
|
|
|
|
"Hungarian",
|
|
|
|
"Icelandic",
|
|
|
|
"Igbo",
|
|
|
|
"Indonesian",
|
|
|
|
"Indonesian (auto-generated)",
|
|
|
|
"Interlingue",
|
|
|
|
"Irish",
|
|
|
|
"Italian",
|
|
|
|
"Italian (auto-generated)",
|
|
|
|
"Japanese",
|
|
|
|
"Japanese (auto-generated)",
|
|
|
|
"Javanese",
|
|
|
|
"Kannada",
|
|
|
|
"Kazakh",
|
|
|
|
"Khmer",
|
|
|
|
"Korean",
|
|
|
|
"Korean (auto-generated)",
|
|
|
|
"Kurdish",
|
|
|
|
"Kyrgyz",
|
|
|
|
"Lao",
|
|
|
|
"Latin",
|
|
|
|
"Latvian",
|
|
|
|
"Lithuanian",
|
|
|
|
"Luxembourgish",
|
|
|
|
"Macedonian",
|
|
|
|
"Malagasy",
|
|
|
|
"Malay",
|
|
|
|
"Malayalam",
|
|
|
|
"Maltese",
|
|
|
|
"Maori",
|
|
|
|
"Marathi",
|
|
|
|
"Mongolian",
|
|
|
|
"Nepali",
|
|
|
|
"Norwegian Bokmål",
|
|
|
|
"Nyanja",
|
|
|
|
"Pashto",
|
|
|
|
"Persian",
|
|
|
|
"Polish",
|
|
|
|
"Portuguese",
|
|
|
|
"Portuguese (auto-generated)",
|
|
|
|
"Portuguese (Brazil)",
|
|
|
|
"Punjabi",
|
|
|
|
"Romanian",
|
|
|
|
"Russian",
|
|
|
|
"Russian (auto-generated)",
|
|
|
|
"Samoan",
|
|
|
|
"Scottish Gaelic",
|
|
|
|
"Serbian",
|
|
|
|
"Shona",
|
|
|
|
"Sindhi",
|
|
|
|
"Sinhala",
|
|
|
|
"Slovak",
|
|
|
|
"Slovenian",
|
|
|
|
"Somali",
|
|
|
|
"Southern Sotho",
|
|
|
|
"Spanish",
|
|
|
|
"Spanish (auto-generated)",
|
|
|
|
"Spanish (Latin America)",
|
|
|
|
"Spanish (Mexico)",
|
|
|
|
"Spanish (Spain)",
|
|
|
|
"Sundanese",
|
|
|
|
"Swahili",
|
|
|
|
"Swedish",
|
|
|
|
"Tajik",
|
|
|
|
"Tamil",
|
|
|
|
"Telugu",
|
|
|
|
"Thai",
|
|
|
|
"Turkish",
|
|
|
|
"Turkish (auto-generated)",
|
|
|
|
"Ukrainian",
|
|
|
|
"Urdu",
|
|
|
|
"Uzbek",
|
|
|
|
"Vietnamese",
|
|
|
|
"Vietnamese (auto-generated)",
|
|
|
|
"Welsh",
|
|
|
|
"Western Frisian",
|
|
|
|
"Xhosa",
|
|
|
|
"Yiddish",
|
|
|
|
"Yoruba",
|
|
|
|
"Zulu",
|
|
|
|
}
|
|
|
|
end
|
|
|
|
end
|