mirror of
https://git.nadeko.net/Fijxu/invidious.git
synced 2026-01-23 15:21:56 +00:00
Captions: Add ability to use Innertube's transcripts API (#4001)
This commit is contained in:
Samantaz Fox
@@ -1,100 +1,106 @@
|
||||
require "json"
|
||||
|
||||
module Invidious::Videos
|
||||
struct Caption
|
||||
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
|
||||
|
||||
def initialize(@name, @language_code, @base_url)
|
||||
end
|
||||
property auto_generated : Bool
|
||||
|
||||
# Parse the JSON structure from Youtube
|
||||
def self.from_yt_json(container : JSON::Any) : Array(Caption)
|
||||
caption_tracks = container
|
||||
.dig?("playerCaptionsTracklistRenderer", "captionTracks")
|
||||
.try &.as_a
|
||||
|
||||
captions_list = [] of Caption
|
||||
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
|
||||
|
||||
captions_list << Caption.new(name, language_code, base_url)
|
||||
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 = (caption["kind"]? == "asr")
|
||||
|
||||
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.
|
||||
|
||||
103
src/invidious/videos/transcript.cr
Normal file
103
src/invidious/videos/transcript.cr
Normal file
@@ -0,0 +1,103 @@
|
||||
module Invidious::Videos
|
||||
# Namespace for methods primarily relating to Transcripts
|
||||
module Transcript
|
||||
record TranscriptLine, start_ms : Time::Span, end_ms : Time::Span, line : String
|
||||
|
||||
def self.generate_param(video_id : String, language_code : String, auto_generated : Bool) : String
|
||||
kind = auto_generated ? "asr" : ""
|
||||
|
||||
object = {
|
||||
"1:0:string" => video_id,
|
||||
|
||||
"2:base64" => {
|
||||
"1:string" => kind,
|
||||
"2:string" => language_code,
|
||||
"3:string" => "",
|
||||
},
|
||||
|
||||
"3:varint" => 1_i64,
|
||||
"5:string" => "engagement-panel-searchable-transcript-search-panel",
|
||||
"6:varint" => 1_i64,
|
||||
"7:varint" => 1_i64,
|
||||
"8:varint" => 1_i64,
|
||||
}
|
||||
|
||||
params = object.try { |i| Protodec::Any.cast_json(i) }
|
||||
.try { |i| Protodec::Any.from_json(i) }
|
||||
.try { |i| Base64.urlsafe_encode(i) }
|
||||
.try { |i| URI.encode_www_form(i) }
|
||||
|
||||
return params
|
||||
end
|
||||
|
||||
def self.convert_transcripts_to_vtt(initial_data : Hash(String, JSON::Any), target_language : String) : String
|
||||
# Convert into array of TranscriptLine
|
||||
lines = self.parse(initial_data)
|
||||
|
||||
# Taken from Invidious::Videos::Captions::Metadata.timedtext_to_vtt()
|
||||
vtt = String.build do |vtt|
|
||||
vtt << <<-END_VTT
|
||||
WEBVTT
|
||||
Kind: captions
|
||||
Language: #{target_language}
|
||||
|
||||
|
||||
END_VTT
|
||||
|
||||
vtt << "\n\n"
|
||||
|
||||
lines.each do |line|
|
||||
start_time = line.start_ms
|
||||
end_time = line.end_ms
|
||||
|
||||
# start_time
|
||||
vtt << start_time.hours.to_s.rjust(2, '0')
|
||||
vtt << ':' << start_time.minutes.to_s.rjust(2, '0')
|
||||
vtt << ':' << start_time.seconds.to_s.rjust(2, '0')
|
||||
vtt << '.' << start_time.milliseconds.to_s.rjust(3, '0')
|
||||
|
||||
vtt << " --> "
|
||||
|
||||
# end_time
|
||||
vtt << end_time.hours.to_s.rjust(2, '0')
|
||||
vtt << ':' << end_time.minutes.to_s.rjust(2, '0')
|
||||
vtt << ':' << end_time.seconds.to_s.rjust(2, '0')
|
||||
vtt << '.' << end_time.milliseconds.to_s.rjust(3, '0')
|
||||
|
||||
vtt << "\n"
|
||||
vtt << line.line
|
||||
|
||||
vtt << "\n"
|
||||
vtt << "\n"
|
||||
end
|
||||
end
|
||||
|
||||
return vtt
|
||||
end
|
||||
|
||||
private def self.parse(initial_data : Hash(String, JSON::Any))
|
||||
body = initial_data.dig("actions", 0, "updateEngagementPanelAction", "content", "transcriptRenderer",
|
||||
"content", "transcriptSearchPanelRenderer", "body", "transcriptSegmentListRenderer",
|
||||
"initialSegments").as_a
|
||||
|
||||
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
|
||||
|
||||
text = extract_text(line["snippet"]) || ""
|
||||
|
||||
lines << TranscriptLine.new(start_ms, end_ms, text)
|
||||
end
|
||||
|
||||
return lines
|
||||
end
|
||||
end
|
||||
end
|
||||
Reference in New Issue
Block a user