ttml, vtt

hyperaudio · Dec 13, 2019 · 4411d07 · 4411d07
1 parent abe3b56
commit 4411d07
Show file tree

Hide file tree

Showing 9 changed files with 164 additions and 91 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1,7 +1,11 @@
 
-input/
-debug/
-output/
+input/*
+debug/*
+output/*
+
+!input/.keep
+!debug/.keep
+!output/.keep
 
 # Created by https://www.gitignore.io/api/node,code,linux,macos,textmate
 # Edit at https://www.gitignore.io/?templates=node,code,linux,macos,textmate

diff --git a/.prettierrc b/.prettierrc
@@ -0,0 +1,4 @@
+singleQuote: true
+trailingComma: es5
+printWidth: 120
+endOfLine: lf
diff --git a/Makefile b/Makefile
@@ -1,4 +1,8 @@
 .PHONY: output clean
+.DEFAULT_GOAL := output
+
+output: input/subs.info.json input/auto-subs.info.json
+	node src/convert
 
 input/subs.info.json:
 	youtube-dl --continue --retries 5 --write-info-json \
@@ -40,9 +44,6 @@ input/auto-subs.info.json:
 		--output "input/auto-subs.%(ext)s" \
 		"https://www.youtube.com/watch?v=hB7aGnfLB-8"
 
-output: input/subs.info.json input/auto-subs.info.json
-	node src/convert
-
 clean:
 	rm -f ./input/*
 	rm -f ./debug/*

diff --git a/debug/.keep b/debug/.keep
diff --git a/input/.keep b/input/.keep
diff --git a/output/.keep b/output/.keep
diff --git a/package.json b/package.json
@@ -10,6 +10,7 @@
   "license": "ISC",
   "dependencies": {
     "shortid": "^2.2.15",
+    "vtt.js": "^0.13.0",
     "xml2js": "^0.4.22"
   }
 }
diff --git a/src/convert.js b/src/convert.js
@@ -1,9 +1,10 @@
-const fs = require("fs");
-const xml2js = require("xml2js");
-const shortid = require("shortid");
+const fs = require('fs');
+const xml2js = require('xml2js');
+const vtt = require('vtt.js');
+const shortid = require('shortid');
 
-const { id: ytid, subtitles } = require("../input/subs.info.json");
-const { automatic_captions } = require("../input/auto-subs.info.json");
+const { id: ytid, subtitles } = require('../input/subs.info.json');
+const { automatic_captions } = require('../input/auto-subs.info.json');
 
 const DEBUG = true;
 
@@ -16,10 +17,32 @@ const generateID = () => {
   return id;
 };
 
+const moreMagic = paragraphs =>
+  paragraphs.reverse().reduce((acc, p) => {
+    if (!p.end && acc.length > 0) p.end = acc[0].start;
+    if (!p.words || p.words.length === 0) {
+      p.words = p.text.split(' ').reduce(
+        (acc, text, index, words) => [
+          ...acc,
+          {
+            id: generateID(),
+            start: p.start + Math.floor((index * (p.end - p.start)) / words.length),
+            end: p.start + Math.floor(((index + 1) * (p.end - p.start)) / words.length),
+            offset: index === 0 ? 0 : acc.map(({ text }) => text).join(' ').length + 1,
+            length: text.length,
+            text,
+          },
+        ],
+        []
+      );
+    }
+    return [p, ...acc];
+  }, []);
+
 const main = () => {
   const queue = [
-    ["subs", subtitles],
-    ["auto-subs", automatic_captions]
+    ['subs', subtitles],
+    ['auto-subs', automatic_captions],
   ]
     .reduce(
       (acc, [type, subs]) => [
@@ -31,128 +54,163 @@ const main = () => {
               lang,
               ext,
               type,
-              file: `${type}.${lang}.${ext}`
-            }))
+              file: `${type}.${lang}.${ext}`,
+            })),
           ],
           []
-        )
+        ),
       ],
       []
     )
     .filter(({ file }) => fs.existsSync(`./input/${file}`));
-  // .filter(({ lang }) => lang === "ro");
+  // .filter(({ lang }) => lang === 'ro');
   // .slice(0, 1);
 
-  console.log(JSON.stringify(queue, null, 2));
+  DEBUG && console.log(JSON.stringify(queue, null, 2));
 
   queue.forEach(({ lang, type, ext, file }) => {
     switch (ext) {
-      case "srv3":
+      case 'srv3':
         convertSRV3(file, lang, type);
         break;
-
+      case 'ttml':
+        convertTTML(file, lang, type);
+        break;
+      case 'vtt':
+        convertVTT(file, lang, type);
+        break;
       default:
         console.warn(`Unknown format ${ext}`);
         break;
     }
   });
 };
 
-const convertSRV3 = async (file, lang, type) => {
-  const data = await xml2js.parseStringPromise(
-    fs.readFileSync(`./input/${file}`, { encoding: "utf-8" }),
+const convertVTT = async (file, lang, type) => {
+  global.navigator = { userAgent: '' };
+  const cues = [];
+
+  const parser = new vtt.WebVTT.Parser(
     {
-      attrkey: "attrs",
-      charkey: "text",
-      trim: true,
-      explicitArray: true
-    }
+      VTTCue: vtt.VTTCue,
+      VTTRegion: vtt.VTTRegion,
+    },
+    vtt.WebVTT.StringDecoder()
   );
 
-  DEBUG &&
-    fs.writeFileSync(
-      `./debug/${file}.parsed.json`,
-      JSON.stringify(data, null, 2),
-      "utf8"
-    );
+  parser.oncue = cue => cues.push(cue);
 
-  const paragraphs = data.timedtext.body[0].p
-    .filter(({ s, text }) => !!s || !!text)
-    .map(({ attrs: { t }, s = [], text }) => {
-      const start = parseInt(t);
+  parser.parse(fs.readFileSync(`./input/${file}`, { encoding: 'utf-8' }));
+  parser.flush();
+
+  DEBUG && fs.writeFileSync(`./debug/${file}.parsed.json`, JSON.stringify(cues, null, 2), 'utf8');
+
+  const paragraphs = moreMagic(
+    cues.map(({ startTime, endTime, text }) => {
+      const start = parseFloat(startTime) * 1e3;
+
+      return {
+        id: generateID(),
+        start,
+        text: text.trim(),
+      };
+    })
+  );
+
+  DEBUG && fs.writeFileSync(`./debug/${file}.paragraphs.json`, JSON.stringify(paragraphs, null, 2), 'utf8');
+
+  const transcript = {
+    id: generateID(),
+    lang,
+    paragraphs,
+  };
+
+  fs.writeFileSync(`./output/${file}.json`, JSON.stringify(transcript, null, 2), 'utf8');
+};
+
+const convertTTML = async (file, lang, type) => {
+  const data = await xml2js.parseStringPromise(fs.readFileSync(`./input/${file}`, { encoding: 'utf-8' }), {
+    attrkey: 'attrs',
+    charkey: 'text',
+    trim: true,
+    explicitArray: true,
+  });
+
+  DEBUG && fs.writeFileSync(`./debug/${file}.parsed.json`, JSON.stringify(data, null, 2), 'utf8');
+
+  const paragraphs = moreMagic(
+    data.tt.body[0].div[0].p.map(({ attrs: { begin }, text }) => {
+      const [hh, mm, ss] = begin.split(':');
+      const start = (hh * 3600 + mm * 60 + ss) * 1e3;
 
-      const words = s.reduce(
-        (acc, { attrs: { t = 0, ac = 0 }, text }, index) => [
-          ...acc,
-          {
-            id: generateID(),
-            start: start + parseInt(t),
-            end: start + parseInt(t) + parseInt(ac),
-            offset:
-              index === 0
-                ? 0
-                : acc.map(({ text }) => text).join(" ").length + 1,
-            length: text.length,
-            text
-          }
-        ],
-        []
-      );
       return {
         id: generateID(),
         start,
-        end: words.length > 0 ? words[words.length - 1].end : null,
-        text: text ? text : words.map(({ text }) => text).join(" "),
-        words
+        text,
       };
     })
-    .reverse()
-    .reduce((acc, p) => {
-      if (!p.end && acc.length > 0) p.end = acc[0].start;
-      if (!p.words || p.words.length === 0) {
-        p.words = p.text.split(" ").reduce(
-          (acc, text, index, words) => [
+  );
+
+  DEBUG && fs.writeFileSync(`./debug/${file}.paragraphs.json`, JSON.stringify(paragraphs, null, 2), 'utf8');
+
+  const transcript = {
+    id: generateID(),
+    lang,
+    paragraphs,
+  };
+
+  fs.writeFileSync(`./output/${file}.json`, JSON.stringify(transcript, null, 2), 'utf8');
+};
+
+const convertSRV3 = async (file, lang, type) => {
+  const data = await xml2js.parseStringPromise(fs.readFileSync(`./input/${file}`, { encoding: 'utf-8' }), {
+    attrkey: 'attrs',
+    charkey: 'text',
+    trim: true,
+    explicitArray: true,
+  });
+
+  DEBUG && fs.writeFileSync(`./debug/${file}.parsed.json`, JSON.stringify(data, null, 2), 'utf8');
+
+  const paragraphs = moreMagic(
+    data.timedtext.body[0].p
+      .filter(({ s, text }) => !!s || !!text)
+      .map(({ attrs: { t }, s = [], text }) => {
+        const start = parseInt(t);
+
+        const words = s.reduce(
+          (acc, { attrs: { t = 0, ac = 0 }, text }, index) => [
             ...acc,
             {
               id: generateID(),
-              start:
-                p.start +
-                Math.floor((index * (p.end - p.start)) / words.length),
-              end:
-                p.start +
-                Math.floor(((index + 1) * (p.end - p.start)) / words.length),
-              offset:
-                index === 0
-                  ? 0
-                  : acc.map(({ text }) => text).join(" ").length + 1,
+              start: start + parseInt(t),
+              end: start + parseInt(t) + parseInt(ac),
+              offset: index === 0 ? 0 : acc.map(({ text }) => text).join(' ').length + 1,
               length: text.length,
-              text
-            }
+              text,
+            },
           ],
           []
         );
-      }
-      return [p, ...acc];
-    }, []);
+        return {
+          id: generateID(),
+          start,
+          end: words.length > 0 ? words[words.length - 1].end : null,
+          text: text ? text : words.map(({ text }) => text).join(' '),
+          words,
+        };
+      })
+  );
 
-  DEBUG &&
-    fs.writeFileSync(
-      `./debug/${file}.paragraphs.json`,
-      JSON.stringify(paragraphs, null, 2),
-      "utf8"
-    );
+  DEBUG && fs.writeFileSync(`./debug/${file}.paragraphs.json`, JSON.stringify(paragraphs, null, 2), 'utf8');
 
   const transcript = {
     id: generateID(),
     lang,
-    paragraphs
+    paragraphs,
   };
 
-  fs.writeFileSync(
-    `./output/${file}.json`,
-    JSON.stringify(transcript, null, 2),
-    "utf8"
-  );
+  fs.writeFileSync(`./output/${file}.json`, JSON.stringify(transcript, null, 2), 'utf8');
 };
 
 main();
diff --git a/yarn.lock b/yarn.lock
@@ -134,6 +134,11 @@ util.promisify@~1.0.0:
     define-properties "^1.1.2"
     object.getownpropertydescriptors "^2.0.3"
 
+vtt.js@^0.13.0:
+  version "0.13.0"
+  resolved "https://registry.yarnpkg.com/vtt.js/-/vtt.js-0.13.0.tgz#955c667b34d5325b2012cb9e8ba9bad6e0b11ff8"
+  integrity sha1-lVxmezTVMlsgEsuei6m61uCxH/g=
+
 xml2js@^0.4.22:
   version "0.4.22"
   resolved "https://registry.yarnpkg.com/xml2js/-/xml2js-0.4.22.tgz#4fa2d846ec803237de86f30aa9b5f70b6600de02"