From 30d32b17017e78db3d9b270b683bfc5ba5dde0ba Mon Sep 17 00:00:00 2001
From: Joey Parrish <joeyparrish@google.com>
Date: Thu, 24 Aug 2017 11:27:03 -0700
Subject: [PATCH] Add support for MP4+TTML in HLS

Previously, our HLS implementation only supported WebVTT, because
that seemed to be the only supported format in HLS.

Now, we add support for other formats by extending the MIME and codec
guessing system to text streams.  We also add a parser registration
for HLS's new "stpp.TTML.im1t" codec string.

Closes #986

Change-Id: I0a58ca4650675ff0546e36edf4114290f7931239
---
 lib/hls/hls_parser.js       | 244 ++++++++++++++++++++++++--------
 lib/text/mp4_ttml_parser.js |   3 +
 test/hls/hls_parser_unit.js | 270 ++++++++++++++++++++++++++++++++++--
 3 files changed, 444 insertions(+), 73 deletions(-)

diff --git a/lib/hls/hls_parser.js b/lib/hls/hls_parser.js
index bc2deb4928..a6c6b1a158 100644
--- a/lib/hls/hls_parser.js
+++ b/lib/hls/hls_parser.js
@@ -60,6 +60,12 @@ shaka.hls.HlsParser = function() {
   /** @private {!Object.<number, shaka.hls.HlsParser.StreamInfo>} */
   this.mediaTagsToStreamInfosMap_ = {};
 
+  /**
+   * The key is a string of the form "<VIDEO URI> - <AUDIO URI>".
+   * @private {!Object.<string, shakaExtern.Variant>}
+   */
+  this.urisToVariantsMap_ = {};
+
   /** @private {!Object.<number, !shaka.media.SegmentIndex>} */
   this.streamsToIndexMap_ = {};
 
@@ -159,6 +165,7 @@ shaka.hls.HlsParser.prototype.stop = function() {
   this.playerInterface_ = null;
   this.config_ = null;
   this.mediaTagsToStreamInfosMap_ = {};
+  this.urisToVariantsMap_ = {};
   this.manifest_ = null;
 
   return Promise.resolve();
@@ -339,12 +346,6 @@ shaka.hls.HlsParser.prototype.createPeriod_ = function(playlist) {
   var Functional = shaka.util.Functional;
   var tags = playlist.tags;
 
-  // Create Variants for every 'EXT-X-STREAM-INF' tag.
-  var variantTags = Utils.filterTagsByName(tags, 'EXT-X-STREAM-INF');
-  var variantsPromises = variantTags.map(function(tag) {
-    return this.createVariantsForTag_(tag, playlist);
-  }.bind(this));
-
   var mediaTags = Utils.filterTagsByName(playlist.tags, 'EXT-X-MEDIA');
   var textStreamTags = mediaTags.filter(function(tag) {
     var type = shaka.hls.HlsParser.getRequiredAttributeValue_(tag, 'TYPE');
@@ -356,8 +357,16 @@ shaka.hls.HlsParser.prototype.createPeriod_ = function(playlist) {
     return this.createTextStream_(tag, playlist);
   }.bind(this));
 
-  return Promise.all(variantsPromises).then(function(allVariants) {
-    return Promise.all(textStreamPromises).then(function(textStreams) {
+  return Promise.all(textStreamPromises).then(function(textStreams) {
+    // Create Variants for every 'EXT-X-STREAM-INF' tag.  Do this after text
+    // streams have been created, so that we can push text codecs found on the
+    // variant tag back into the created text streams.
+    var variantTags = Utils.filterTagsByName(tags, 'EXT-X-STREAM-INF');
+    var variantsPromises = variantTags.map(function(tag) {
+      return this.createVariantsForTag_(tag, playlist);
+    }.bind(this));
+
+    return Promise.all(variantsPromises).then(function(allVariants) {
       var variants = allVariants.reduce(Functional.collapseArrays, []);
       if (!this.isLive_)
         this.fitSegments_(variants);
@@ -383,8 +392,6 @@ shaka.hls.HlsParser.prototype.createVariantsForTag_ = function(tag, playlist) {
   var ContentType = shaka.util.ManifestParserUtils.ContentType;
   var HlsParser = shaka.hls.HlsParser;
   var Utils = shaka.hls.Utils;
-  var bandwidth =
-      Number(HlsParser.getRequiredAttributeValue_(tag, 'BANDWIDTH'));
 
   // These are the default codecs to assume if none are specified.
   //
@@ -400,6 +407,8 @@ shaka.hls.HlsParser.prototype.createVariantsForTag_ = function(tag, playlist) {
   var width = null;
   var height = null;
   var frameRate = tag.getAttributeValue('FRAME-RATE');
+  var bandwidth =
+      Number(HlsParser.getRequiredAttributeValue_(tag, 'BANDWIDTH'));
 
   if (resolutionAttr) {
     var resBlocks = resolutionAttr.value.split('x');
@@ -409,6 +418,8 @@ shaka.hls.HlsParser.prototype.createVariantsForTag_ = function(tag, playlist) {
 
   var timeOffset = this.getTimeOffset_(playlist);
 
+  // After filtering, this is a list of the media tags we will process to
+  // combine with the variant tag (EXT-X-STREAM-INF) we are working on.
   var mediaTags = Utils.filterTagsByName(playlist.tags, 'EXT-X-MEDIA');
 
   var audioGroupId = tag.getAttributeValue('AUDIO');
@@ -422,6 +433,30 @@ shaka.hls.HlsParser.prototype.createVariantsForTag_ = function(tag, playlist) {
   } else if (videoGroupId) {
     mediaTags = Utils.findMediaTags(mediaTags, 'VIDEO', videoGroupId);
   }
+
+  // There may be a codec string for the text stream.  We should identify it,
+  // add it to the appropriate stream, then strip it out of the variant to
+  // avoid confusing our multiplex detection below.
+  var textCodecs = this.guessCodecsSafe_(ContentType.TEXT, codecs);
+  if (textCodecs) {
+    // We found a text codec in the list, so look for an associated text stream.
+    var subGroupId = tag.getAttributeValue('SUBTITLES');
+    if (subGroupId) {
+      var textTags = Utils.findMediaTags(mediaTags, 'SUBTITLES', subGroupId);
+      goog.asserts.assert(textTags.length == 1,
+                          'Exactly one text tag expected!');
+      if (textTags.length) {
+        // We found a text codec and text stream, so make sure the codec is
+        // attached to the stream.
+        var textStreamInfo = this.mediaTagsToStreamInfosMap_[textTags[0].id];
+        textStreamInfo.stream.codecs = textCodecs;
+      }
+    }
+
+    // Remove this entry from the list of codecs that belong to audio/video.
+    codecs.splice(codecs.indexOf(textCodecs), 1);
+  }
+
   var promises = mediaTags.map(function(tag) {
     return this.createStreamInfoFromMediaTag_(tag, codecs, timeOffset);
   }.bind(this));
@@ -446,7 +481,9 @@ shaka.hls.HlsParser.prototype.createVariantsForTag_ = function(tag, playlist) {
 
       if (codecs.length == 1) {
         // There is only one codec, so it shouldn't be multiplexed.
-        if (resolutionAttr || frameRate) {
+
+        var videoCodecs = this.guessCodecsSafe_(ContentType.VIDEO, codecs);
+        if (resolutionAttr || frameRate || videoCodecs) {
           // Assume video-only.
           shaka.log.debug('Guessing video-only.');
           type = ContentType.VIDEO;
@@ -565,8 +602,19 @@ shaka.hls.HlsParser.prototype.createVariants_ =
         drmInfos = videoDrmInfos;
       }
 
-      variants.push(this.createVariant_(
-          audioStream, videoStream, bandwidth, drmInfos));
+      var videoStreamUri = videoInfos[i] ? videoInfos[i].relativeUri : '';
+      var audioStreamUri = audioInfos[i] ? audioInfos[i].relativeUri : '';
+      var variantMapKey = videoStreamUri + ' - ' + audioStreamUri;
+      if (this.urisToVariantsMap_[variantMapKey]) {
+        // This happens when two variants only differ in their text streams.
+        shaka.log.debug('Skipping variant which only differs in text streams.');
+        continue;
+      }
+
+      var variant = this.createVariant_(
+          audioStream, videoStream, bandwidth, drmInfos);
+      variants.push(variant);
+      this.urisToVariantsMap_[variantMapKey] = variant;
     }
   }
   return variants;
@@ -859,7 +907,7 @@ shaka.hls.HlsParser.prototype.createStreamInfo_ = function(uri, allCodecs,
           shaka.util.Error.Code.HLS_KEYFORMATS_NOT_SUPPORTED);
     }
 
-    return this.guessMimeType_(type, segments[0].getUris()[0])
+    return this.guessMimeType_(type, codecs, segments[0].getUris()[0])
         .then(function(mimeType) {
           var stream = {
             id: this.globalId_++,
@@ -1097,29 +1145,19 @@ shaka.hls.HlsParser.prototype.fitSegments_ = function(variants) {
 
 
 /**
- * Attempts to guess which codecs from the codecs list belong
- * to a given content type.
+ * Attempts to guess which codecs from the codecs list belong to a given content
+ * type.  Does not assume a single codec is anything special, and does not throw
+ * if it fails to match.
  *
  * @param {!string} contentType
  * @param {!Array.<!string>} codecs
- * @return {string}
+ * @return {?string} or null if no match is found
  * @private
- * @throws {shaka.util.Error}
  */
-shaka.hls.HlsParser.prototype.guessCodecs_ = function(contentType, codecs) {
-  if (codecs.length == 1) {
-    return codecs[0];
-  }
-
+shaka.hls.HlsParser.prototype.guessCodecsSafe_ = function(contentType, codecs) {
   var ContentType = shaka.util.ManifestParserUtils.ContentType;
-  if (contentType == ContentType.TEXT) {
-    return '';
-  }
-
   var HlsParser = shaka.hls.HlsParser;
-  var formats = HlsParser.VIDEO_CODEC_FORMATS_;
-  if (contentType == ContentType.AUDIO)
-    formats = HlsParser.AUDIO_CODEC_FORMATS_;
+  var formats = HlsParser.CODEC_REGEXPS_BY_CONTENT_TYPE_[contentType];
 
   for (var i = 0; i < formats.length; i++) {
     for (var j = 0; j < codecs.length; j++) {
@@ -1129,6 +1167,35 @@ shaka.hls.HlsParser.prototype.guessCodecs_ = function(contentType, codecs) {
     }
   }
 
+  // Text does not require a codec string.
+  if (contentType == ContentType.TEXT) {
+    return '';
+  }
+
+  return null;
+};
+
+
+/**
+ * Attempts to guess which codecs from the codecs list belong to a given content
+ * type.  Assumes a single codec is correct, and throws if not found.
+ *
+ * @param {!string} contentType
+ * @param {!Array.<!string>} codecs
+ * @return {string}
+ * @private
+ * @throws {shaka.util.Error}
+ */
+shaka.hls.HlsParser.prototype.guessCodecs_ = function(contentType, codecs) {
+  if (codecs.length == 1) {
+    return codecs[0];
+  }
+
+  var match = this.guessCodecsSafe_(contentType, codecs);
+  if (match != null) {
+    return match;
+  }
+
   // Unable to guess codecs.
   throw new shaka.util.Error(
       shaka.util.Error.Severity.CRITICAL,
@@ -1142,30 +1209,38 @@ shaka.hls.HlsParser.prototype.guessCodecs_ = function(contentType, codecs) {
  * Attempts to guess stream's mime type based on content type and uri.
  *
  * @param {!string} contentType
+ * @param {!string} codecs
  * @param {!string} uri
  * @return {!Promise.<!string>}
  * @private
  * @throws {shaka.util.Error}
  */
-shaka.hls.HlsParser.prototype.guessMimeType_ = function(contentType, uri) {
+shaka.hls.HlsParser.prototype.guessMimeType_ =
+    function(contentType, codecs, uri) {
   var ContentType = shaka.util.ManifestParserUtils.ContentType;
+  var HlsParser = shaka.hls.HlsParser;
+
   var blocks = uri.split('.');
   var extension = blocks[blocks.length - 1];
-
-  if (contentType == ContentType.TEXT) {
-    // HLS only supports vtt at the moment.
-    return Promise.resolve('text/vtt');
-  }
-
-  var HlsParser = shaka.hls.HlsParser;
-  var map = HlsParser.AUDIO_EXTENSIONS_TO_MIME_TYPES_;
-  if (contentType == ContentType.VIDEO)
-    map = HlsParser.VIDEO_EXTENSIONS_TO_MIME_TYPES_;
+  var map = HlsParser.EXTENSION_MAP_BY_CONTENT_TYPE_[contentType];
 
   var mimeType = map[extension];
   if (mimeType)
     return Promise.resolve(mimeType);
 
+  if (contentType == ContentType.TEXT) {
+    // The extension map didn't work.
+    if (!codecs || codecs == 'vtt') {
+      // If codecs is 'vtt', it's WebVTT.
+      // If there was no codecs string, assume HLS text streams are WebVTT.
+      return Promise.resolve('text/vtt');
+    } else {
+      // Otherwise, assume MP4-embedded text, since text-based formats tend not
+      // to have a codecs string at all.
+      return Promise.resolve('application/mp4');
+    }
+  }
+
   // If unable to guess mime type, request a segment and try getting it
   // from the response.
   var headRequest = shaka.net.NetworkingEngine.makeRequest(
@@ -1184,7 +1259,8 @@ shaka.hls.HlsParser.prototype.guessMimeType_ = function(contentType, uri) {
               extension);
         }
 
-        return mimeType;
+        // Split the MIME type in case the server sent additional parameters.
+        return mimeType.split(';')[0];
       });
 };
 
@@ -1293,37 +1369,61 @@ shaka.hls.HlsParser.prototype.requestManifest_ = function(uri) {
 
 
 /**
- * A list of well-known video codecs formats.
+ * A list of regexps to detect well-known video codecs.
  *
- * @const {!Array<!RegExp>}
+ * @const {!Array.<!RegExp>}
  * @private
  */
-shaka.hls.HlsParser.VIDEO_CODEC_FORMATS_ = [
-  /^(avc)/,
-  /^(hvc)/,
-  /^(vp[8-9])$/,
-  /^(av1)$/,
-  /^(mp4v)/
+shaka.hls.HlsParser.VIDEO_CODEC_REGEXPS_ = [
+  /^avc/,
+  /^hev/,
+  /^hvc/,
+  /^vp0?[89]/,
+  /^av1$/
 ];
 
 
 /**
- * A list of well-known audio codecs formats.
+ * A list of regexps to detect well-known audio codecs.
  *
- * @const {!Array<!RegExp>}
+ * @const {!Array.<!RegExp>}
  * @private
  */
-shaka.hls.HlsParser.AUDIO_CODEC_FORMATS_ = [
-  /^(vorbis)/,
-  /^(opus)/,
-  /^(mp4a)/,
-  /^(ac-3)$/,
-  /^(ec-3)$/
+shaka.hls.HlsParser.AUDIO_CODEC_REGEXPS_ = [
+  /^vorbis$/,
+  /^opus$/,
+  /^flac$/,
+  /^mp4a/,
+  /^[ae]c-3$/
 ];
 
 
 /**
- * @const {!Object<string, string>}
+ * A list of regexps to detect well-known text codecs.
+ *
+ * @const {!Array.<!RegExp>}
+ * @private
+ */
+shaka.hls.HlsParser.TEXT_CODEC_REGEXPS_ = [
+  /^vtt$/,
+  /^wvtt/,
+  /^stpp/
+];
+
+
+/**
+ * @const {!Object.<string, !Array.<!RegExp>>}
+ * @private
+ */
+shaka.hls.HlsParser.CODEC_REGEXPS_BY_CONTENT_TYPE_ = {
+  'audio': shaka.hls.HlsParser.AUDIO_CODEC_REGEXPS_,
+  'video': shaka.hls.HlsParser.VIDEO_CODEC_REGEXPS_,
+  'text': shaka.hls.HlsParser.TEXT_CODEC_REGEXPS_
+};
+
+
+/**
+ * @const {!Object.<string, string>}
  * @private
  */
 shaka.hls.HlsParser.AUDIO_EXTENSIONS_TO_MIME_TYPES_ = {
@@ -1337,7 +1437,7 @@ shaka.hls.HlsParser.AUDIO_EXTENSIONS_TO_MIME_TYPES_ = {
 
 
 /**
- * @const {!Object<string, string>}
+ * @const {!Object.<string, string>}
  * @private
  */
 shaka.hls.HlsParser.VIDEO_EXTENSIONS_TO_MIME_TYPES_ = {
@@ -1349,6 +1449,30 @@ shaka.hls.HlsParser.VIDEO_EXTENSIONS_TO_MIME_TYPES_ = {
 };
 
 
+/**
+ * @const {!Object.<string, string>}
+ * @private
+ */
+shaka.hls.HlsParser.TEXT_EXTENSIONS_TO_MIME_TYPES_ = {
+  'mp4': 'application/mp4',
+  'm4s': 'application/mp4',
+  'm4i': 'application/mp4',
+  'vtt': 'text/vtt',
+  'ttml': 'application/ttml+xml'
+};
+
+
+/**
+ * @const {!Object.<string, !Object.<string, string>>}
+ * @private
+ */
+shaka.hls.HlsParser.EXTENSION_MAP_BY_CONTENT_TYPE_ = {
+  'audio': shaka.hls.HlsParser.AUDIO_EXTENSIONS_TO_MIME_TYPES_,
+  'video': shaka.hls.HlsParser.VIDEO_EXTENSIONS_TO_MIME_TYPES_,
+  'text': shaka.hls.HlsParser.TEXT_EXTENSIONS_TO_MIME_TYPES_
+};
+
+
 /**
  * @typedef {function(!shaka.hls.Tag):?shakaExtern.DrmInfo}
  * @private
diff --git a/lib/text/mp4_ttml_parser.js b/lib/text/mp4_ttml_parser.js
index ab1b2a34f5..4bb756b632 100644
--- a/lib/text/mp4_ttml_parser.js
+++ b/lib/text/mp4_ttml_parser.js
@@ -91,3 +91,6 @@ shaka.text.Mp4TtmlParser.prototype.parseMedia = function(data, time) {
 shaka.text.TextEngine.registerParser(
     'application/mp4; codecs="stpp"',
     shaka.text.Mp4TtmlParser);
+shaka.text.TextEngine.registerParser(
+    'application/mp4; codecs="stpp.TTML.im1t"',
+    shaka.text.Mp4TtmlParser);
diff --git a/test/hls/hls_parser_unit.js b/test/hls/hls_parser_unit.js
index b59ce3d39a..4e2dc7b294 100644
--- a/test/hls/hls_parser_unit.js
+++ b/test/hls/hls_parser_unit.js
@@ -27,6 +27,8 @@ describe('HlsParser', function() {
   var playerInterface;
   /** @type {shakaExtern.ManifestConfiguration} */
   var config;
+  /** @const */
+  var TextStreamKind = shaka.util.ManifestParserUtils.TextStreamKind;
 
   beforeEach(function() {
     fakeNetEngine = new shaka.test.FakeNetworkingEngine();
@@ -65,8 +67,9 @@ describe('HlsParser', function() {
     fakeNetEngine.setResponseMapAsText({
       'test://master': master,
       'test://audio': media,
+      'test://audio2': media,
       'test://video': media,
-      'test://text': media
+      'test://video2': media
     });
 
     parser.start('test://master', playerInterface)
@@ -110,6 +113,38 @@ describe('HlsParser', function() {
     testHlsParser(master, media, manifest, done);
   });
 
+  it('guesses video-only variant by codecs', function(done) {
+    var master = [
+      '#EXTM3U\n',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1"\n',
+      'test://video'
+    ].join('');
+
+    var media = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXT-X-MAP:URI="test://main.mp4",BYTERANGE="616@0"\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.mp4'
+    ].join('');
+
+    var manifest = new shaka.test.ManifestGenerator()
+            .anyTimeline()
+            .addPeriod(jasmine.any(Number))
+              .addVariant(jasmine.any(Number))
+                .language('und')
+                .bandwidth(200)
+                .addVideo(jasmine.any(Number))
+                  .anySegmentFunctions()
+                  .anyInitSegment()
+                  .presentationTimeOffset(0)
+                  .mime('video/mp4', 'avc1')
+          .build();
+
+    testHlsParser(master, media, manifest, done);
+  });
+
   it('parses audio-only variant', function(done) {
     var master = [
       '#EXTM3U\n',
@@ -145,7 +180,7 @@ describe('HlsParser', function() {
   it('parses audio+video variant', function(done) {
     var master = [
       '#EXTM3U\n',
-      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1, mp4a",',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,mp4a",',
       'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1"\n',
       'test://video\n',
       '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud1",LANGUAGE="eng",',
@@ -337,16 +372,16 @@ describe('HlsParser', function() {
   it('parses multiple variants', function(done) {
     var master = [
       '#EXTM3U\n',
-      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1, mp4a",',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,mp4a",',
       'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1"\n',
       'test://video\n',
       '#EXT-X-STREAM-INF:BANDWIDTH=300,CODECS="avc1,mp4a",',
       'RESOLUTION=960x540,FRAME-RATE=120,AUDIO="aud2"\n',
-      'test://video\n',
+      'test://video2\n',
       '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud1",LANGUAGE="eng",',
       'URI="test://audio"\n',
       '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud2",LANGUAGE="fr",',
-      'URI="test://audio"\n'
+      'URI="test://audio2"\n'
     ].join('');
 
     var media = [
@@ -479,8 +514,7 @@ describe('HlsParser', function() {
     fakeNetEngine.setResponseMapAsText({
       'test://master': master,
       'test://audio': media,
-      'test://video': media,
-      'test://text': media
+      'test://video': media
     });
 
     var filterAllPeriods = jasmine.createSpy('filterAllPeriods');
@@ -524,7 +558,8 @@ describe('HlsParser', function() {
                   .size(960, 540)
           .build();
 
-    var headers = {'content-type': 'video/mp4'};
+    // The extra parameters should be stripped by the parser.
+    var headers = {'content-type': 'video/mp4; foo=bar'};
     fakeNetEngine.setHeadersMap({
       'test://main.test': headers
     });
@@ -535,15 +570,96 @@ describe('HlsParser', function() {
   it('parses manifest with text streams', function(done) {
     var master = [
       '#EXTM3U\n',
+      '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud1",LANGUAGE="eng",',
+      'URI="test://audio"\n',
+      '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub1",LANGUAGE="eng",',
+      'URI="test://text"\n',
+      '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub2",LANGUAGE="es",',
+      'URI="test://text"\n',
       '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,mp4a",',
-      'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1"\n',
+      'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1",SUBTITLES="sub1"\n',
       'test://video\n',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,mp4a",',
+      'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1",SUBTITLES="sub2"\n',
+      'test://video\n'
+    ].join('');
+
+    var media = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXT-X-MAP:URI="test://main.mp4",BYTERANGE="616@0"\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.mp4'
+    ].join('');
+
+    var textMedia = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.vtt'
+    ].join('');
+
+    var manifest = new shaka.test.ManifestGenerator()
+            .anyTimeline()
+            .addPeriod(jasmine.any(Number))
+              .addVariant(jasmine.any(Number))
+                .language('en')
+                .bandwidth(200)
+                .addVideo(jasmine.any(Number))
+                  .anySegmentFunctions()
+                  .anyInitSegment()
+                  .presentationTimeOffset(0)
+                  .mime('video/mp4', 'avc1')
+                  .frameRate(60)
+                  .size(960, 540)
+                .addAudio(jasmine.any(Number))
+                  .language('en')
+                  .anySegmentFunctions()
+                  .anyInitSegment()
+                  .presentationTimeOffset(0)
+                  .mime('audio/mp4', 'mp4a')
+              .addTextStream(jasmine.any(Number))
+                .language('en')
+                .anySegmentFunctions()
+                .nullInitSegment()
+                .presentationTimeOffset(0)
+                .mime('text/vtt', '')
+                .kind(TextStreamKind.SUBTITLE)
+              .addTextStream(jasmine.any(Number))
+                .language('es')
+                .anySegmentFunctions()
+                .nullInitSegment()
+                .presentationTimeOffset(0)
+                .mime('text/vtt', '')
+                .kind(TextStreamKind.SUBTITLE)
+          .build();
+
+    fakeNetEngine.setResponseMapAsText({
+      'test://master': master,
+      'test://audio': media,
+      'test://video': media,
+      'test://text': textMedia
+    });
+
+    parser.start('test://master', playerInterface)
+        .then(function(actual) { expect(actual).toEqual(manifest); })
+        .catch(fail).then(done);
+  });
+
+  it('parses manifest with text streams without SUBTITLES', function(done) {
+    var master = [
+      '#EXTM3U\n',
       '#EXT-X-MEDIA:TYPE=AUDIO,GROUP-ID="aud1",LANGUAGE="eng",',
       'URI="test://audio"\n',
       '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub1",LANGUAGE="eng",',
       'URI="test://text"\n',
       '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub2",LANGUAGE="es",',
-      'URI="test://text"\n'
+      'URI="test://text"\n',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,mp4a",',
+      'RESOLUTION=960x540,FRAME-RATE=60,AUDIO="aud1"\n',
+      'test://video\n'
     ].join('');
 
     var media = [
@@ -555,7 +671,14 @@ describe('HlsParser', function() {
       'test://main.mp4'
     ].join('');
 
-    var TextStreamKind = shaka.util.ManifestParserUtils.TextStreamKind;
+    var textMedia = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.vtt'
+    ].join('');
+
     var manifest = new shaka.test.ManifestGenerator()
             .anyTimeline()
             .addPeriod(jasmine.any(Number))
@@ -588,10 +711,131 @@ describe('HlsParser', function() {
                 .nullInitSegment()
                 .presentationTimeOffset(0)
                 .mime('text/vtt', '')
-                .kind('subtitle')
+                .kind(TextStreamKind.SUBTITLE)
           .build();
 
-    testHlsParser(master, media, manifest, done);
+    fakeNetEngine.setResponseMapAsText({
+      'test://master': master,
+      'test://audio': media,
+      'test://video': media,
+      'test://text': textMedia
+    });
+
+    parser.start('test://master', playerInterface)
+        .then(function(actual) { expect(actual).toEqual(manifest); })
+        .catch(fail).then(done);
+  });
+
+  it('parses manifest with MP4+TTML streams', function(done) {
+    var master = [
+      '#EXTM3U\n',
+      '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub1",LANGUAGE="eng",',
+      'URI="test://text"\n',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,stpp.TTML.im1t",',
+      'RESOLUTION=960x540,FRAME-RATE=60,SUBTITLES="sub1"\n',
+      'test://video\n'
+    ].join('');
+
+    var media = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXT-X-MAP:URI="test://main.mp4",BYTERANGE="616@0"\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.mp4'
+    ].join('');
+
+    var manifest = new shaka.test.ManifestGenerator()
+            .anyTimeline()
+            .addPeriod(jasmine.any(Number))
+              .addVariant(jasmine.any(Number))
+                .bandwidth(200)
+                .addVideo(jasmine.any(Number))
+                  .anySegmentFunctions()
+                  .anyInitSegment()
+                  .presentationTimeOffset(0)
+                  .mime('video/mp4', 'avc1')
+                  .frameRate(60)
+                  .size(960, 540)
+              .addTextStream(jasmine.any(Number))
+                .language('en')
+                .anySegmentFunctions()
+                .nullInitSegment()
+                .presentationTimeOffset(0)
+                .mime('application/mp4', 'stpp.TTML.im1t')
+                .kind(TextStreamKind.SUBTITLE)
+          .build();
+
+    fakeNetEngine.setResponseMapAsText({
+      'test://master': master,
+      'test://audio': media,
+      'test://video': media,
+      'test://text': media
+    });
+
+    parser.start('test://master', playerInterface)
+        .then(function(actual) { expect(actual).toEqual(manifest); })
+        .catch(fail).then(done);
+  });
+
+  it('detects VTT streams by codec', function(done) {
+    var master = [
+      '#EXTM3U\n',
+      '#EXT-X-MEDIA:TYPE=SUBTITLES,GROUP-ID="sub1",LANGUAGE="eng",',
+      'URI="test://text"\n',
+      '#EXT-X-STREAM-INF:BANDWIDTH=200,CODECS="avc1,vtt",',
+      'RESOLUTION=960x540,FRAME-RATE=60,SUBTITLES="sub1"\n',
+      'test://video\n'
+    ].join('');
+
+    var media = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXT-X-MAP:URI="test://main.mp4",BYTERANGE="616@0"\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.mp4'
+    ].join('');
+
+    var textMedia = [
+      '#EXTM3U\n',
+      '#EXT-X-PLAYLIST-TYPE:VOD\n',
+      '#EXTINF:5,\n',
+      '#EXT-X-BYTERANGE:121090@616\n',
+      'test://main.foo'
+    ].join('');
+
+    var manifest = new shaka.test.ManifestGenerator()
+            .anyTimeline()
+            .addPeriod(jasmine.any(Number))
+              .addVariant(jasmine.any(Number))
+                .bandwidth(200)
+                .addVideo(jasmine.any(Number))
+                  .anySegmentFunctions()
+                  .anyInitSegment()
+                  .presentationTimeOffset(0)
+                  .mime('video/mp4', 'avc1')
+                  .frameRate(60)
+                  .size(960, 540)
+              .addTextStream(jasmine.any(Number))
+                .language('en')
+                .anySegmentFunctions()
+                .nullInitSegment()
+                .presentationTimeOffset(0)
+                .mime('text/vtt', 'vtt')
+                .kind(TextStreamKind.SUBTITLE)
+          .build();
+
+    fakeNetEngine.setResponseMapAsText({
+      'test://master': master,
+      'test://audio': media,
+      'test://video': media,
+      'test://text': textMedia
+    });
+
+    parser.start('test://master', playerInterface)
+        .then(function(actual) { expect(actual).toEqual(manifest); })
+        .catch(fail).then(done);
   });
 
   it('respects config.hls.defaultTimeOffset setting', function(done) {