feat(toDash): Add support for stable volume/DRC (#662)

LuanRT · May 28, 2024 · 031ffb6 · 031ffb6
1 parent 8e942ad
commit 031ffb6
Show file tree

Hide file tree

Showing 5 changed files with 114 additions and 27 deletions.
diff --git a/src/core/mixins/MediaInfo.ts b/src/core/mixins/MediaInfo.ts
@@ -59,7 +59,17 @@ export default class MediaInfo {
       storyboards = player_response.storyboards;
     }
 
-    return FormatUtils.toDash(this.streaming_data, this.page[0].video_details?.is_post_live_dvr, url_transformer, format_filter, this.#cpn, this.#actions.session.player, this.#actions, storyboards);
+    return FormatUtils.toDash(
+      this.streaming_data,
+      this.page[0].video_details?.is_post_live_dvr,
+      url_transformer,
+      format_filter,
+      this.#cpn,
+      this.#actions.session.player,
+      this.#actions,
+      storyboards,
+      options
+    );
   }
 
   /**

diff --git a/src/types/DashOptions.ts b/src/types/DashOptions.ts
@@ -1,4 +1,6 @@
-export interface DashOptions {
+import type { StreamingInfoOptions } from './StreamingInfoOptions.js';
+
+export interface DashOptions extends StreamingInfoOptions {
   /**
    * Include the storyboards in the DASH manifest when YouTube provides them.
    * Not all players support parsing and displaying storyboards.

diff --git a/src/types/StreamingInfoOptions.ts b/src/types/StreamingInfoOptions.ts
@@ -0,0 +1,21 @@
+export interface StreamingInfoOptions {
+  /**
+   * The label to use for the non-DRC streams when a video has DRC and streams.
+   *
+   * Defaults to `"Original"`
+   */
+  label_original?: string;
+  /**
+   * The label to use for the DRC streams when a video has DRC streams.
+   *
+   * Defaults to `"Stable Volume"`
+   */
+  label_drc?: string;
+  /**
+   * A function that generates the label to use for the DRC streams when a video has multiple audio tracks and DRC streams.
+   * The non-DRC streams use the unmodified audio track label provided by YouTube.
+   *
+   * Defaults to `(audio_track_display_name) => audio_track_display_name + " (Stable Volume)"`
+   */
+  label_drc_mutiple?: (audio_track_display_name: string) => string;
+}
diff --git a/src/utils/DashManifest.tsx b/src/utils/DashManifest.tsx
@@ -12,12 +12,14 @@ import type { PlayerStoryboardSpec } from '../parser/nodes.js';
 import type { SegmentInfo as FSegmentInfo } from './StreamingInfo.js';
 import type { FormatFilter, URLTransformer } from '../types/FormatUtils.js';
 import type PlayerLiveStoryboardSpec from '../parser/classes/PlayerLiveStoryboardSpec.js';
+import type { StreamingInfoOptions } from '../types/StreamingInfoOptions.js';
 
 interface DashManifestProps {
   streamingData: IStreamingData;
   isPostLiveDvr: boolean;
   transformURL?: URLTransformer;
   rejectFormat?: FormatFilter;
+  options?: StreamingInfoOptions,
   cpn?: string;
   player?: Player;
   actions?: Actions;
@@ -70,14 +72,15 @@ async function DashManifest({
   cpn,
   player,
   actions,
-  storyboards
+  storyboards,
+  options
 }: DashManifestProps) {
   const {
     getDuration,
     audio_sets,
     video_sets,
     image_sets
-  } = getStreamingInfo(streamingData, isPostLiveDvr, transformURL, rejectFormat, cpn, player, actions, storyboards);
+  } = getStreamingInfo(streamingData, isPostLiveDvr, transformURL, rejectFormat, cpn, player, actions, storyboards, options);
 
   // XXX: DASH spec: https://standards.iso.org/ittf/PubliclyAvailableStandards/c083314_ISO_IEC%2023009-1_2022(en).zip
 
@@ -104,11 +107,12 @@ async function DashManifest({
             contentType="audio"
           >
             {
-              set.track_role &&
-              <role
-                schemeIdUri="urn:mpeg:dash:role:2011"
-                value={set.track_role}
-              />
+              set.track_roles && set.track_roles.map((role) => (
+                <role
+                  schemeIdUri="urn:mpeg:dash:role:2011"
+                  value={role}
+                />
+              ))
             }
             {
               set.track_name &&
@@ -237,7 +241,8 @@ export function toDash(
   cpn?: string,
   player?: Player,
   actions?: Actions,
-  storyboards?: PlayerStoryboardSpec | PlayerLiveStoryboardSpec
+  storyboards?: PlayerStoryboardSpec | PlayerLiveStoryboardSpec,
+  options?: StreamingInfoOptions
 ) {
   if (!streaming_data)
     throw new InnertubeError('Streaming data not available');
@@ -247,6 +252,7 @@ export function toDash(
       streamingData={streaming_data}
       isPostLiveDvr={is_post_live_dvr}
       transformURL={url_transformer}
+      options={options}
       rejectFormat={format_filter}
       cpn={cpn}
       player={player}

diff --git a/src/utils/StreamingInfo.ts b/src/utils/StreamingInfo.ts
@@ -11,6 +11,7 @@ import type { IStreamingData } from '../parser/index.js';
 import type { Format } from '../parser/misc.js';
 import type { PlayerLiveStoryboardSpec } from '../parser/nodes.js';
 import type { FormatFilter, URLTransformer } from '../types/FormatUtils.js';
+import type { StreamingInfoOptions } from '../types/StreamingInfoOptions.js';
 
 const TAG_ = 'StreamingInfo';
 
@@ -27,7 +28,7 @@ export interface AudioSet {
   codecs?: string;
   audio_sample_rate?: number;
   track_name?: string;
-  track_role?: 'main' | 'dub' | 'description' | 'alternate';
+  track_roles?: ('main' | 'dub' | 'description' | 'enhanced-audio-intelligibility' | 'alternate')[];
   channels?: number;
   representations: AudioRepresentation[];
 }
@@ -130,6 +131,12 @@ interface SharedPostLiveDvrInfo {
   item?: PostLiveDvrInfo
 }
 
+interface DrcLabels {
+  label_original: string;
+  label_drc: string;
+  label_drc_mutiple: (audio_track_display_name: string) => string;
+}
+
 function getFormatGroupings(formats: Format[], is_post_live_dvr: boolean) {
   const group_info = new Map<string, Format[]>();
 
@@ -149,7 +156,9 @@ function getFormatGroupings(formats: Format[], is_post_live_dvr: boolean) {
 
     const audio_track_id = format.audio_track?.id || '';
 
-    const group_id = `${mime_type}-${just_codec}-${color_info}-${audio_track_id}`;
+    const drc = format.is_drc ? 'drc' : '';
+
+    const group_id = `${mime_type}-${just_codec}-${color_info}-${audio_track_id}-${drc}`;
 
     if (!group_info.has(group_id)) {
       group_info.set(group_id, []);
@@ -373,8 +382,18 @@ function getAudioRepresentation(
   const url = new URL(format.decipher(player));
   url.searchParams.set('cpn', cpn || '');
 
+  const uid_parts = [ format.itag.toString() ];
+
+  if (format.audio_track) {
+    uid_parts.push(format.audio_track.id);
+  }
+
+  if (format.is_drc) {
+    uid_parts.push('drc');
+  }
+
   const rep: AudioRepresentation = {
-    uid: format.audio_track ? `${format.itag}-${format.audio_track.id}` : format.itag.toString(),
+    uid: uid_parts.join('-'),
     bitrate: format.bitrate,
     codecs: !hoisted.includes('codecs') ? getStringBetweenStrings(format.mime_type, 'codecs="', '"') : undefined,
     audio_sample_rate: !hoisted.includes('audio_sample_rate') ? format.audio_sample_rate : undefined,
@@ -385,22 +404,25 @@ function getAudioRepresentation(
   return rep;
 }
 
-function getTrackRole(format: Format) {
-  const { audio_track } = format;
-
-  if (!audio_track)
+function getTrackRoles(format: Format, has_drc_streams: boolean) {
+  if (!format.audio_track && !has_drc_streams) {
     return;
+  }
 
-  if (audio_track.audio_is_default)
-    return 'main';
+  const roles: ('main' | 'dub' | 'description' | 'enhanced-audio-intelligibility' | 'alternate')[] = [
+    format.is_original ? 'main' : 'alternate'
+  ];
 
   if (format.is_dubbed)
-    return 'dub';
+    roles.push('dub');
 
   if (format.is_descriptive)
-    return 'description';
+    roles.push('description');
+
+  if (format.is_drc)
+    roles.push('enhanced-audio-intelligibility');
 
-  return 'alternate';
+  return roles;
 }
 
 function getAudioSet(
@@ -409,19 +431,34 @@ function getAudioSet(
   actions?: Actions,
   player?: Player,
   cpn?: string,
-  shared_post_live_dvr_info?: SharedPostLiveDvrInfo
+  shared_post_live_dvr_info?: SharedPostLiveDvrInfo,
+  drc_labels?: DrcLabels
 ) {
   const first_format = formats[0];
   const { audio_track } = first_format;
   const hoisted: string[] = [];
 
+  const has_drc_streams = !!drc_labels;
+
+  let track_name;
+
+  if (audio_track) {
+    if (has_drc_streams && first_format.is_drc) {
+      track_name = drc_labels.label_drc_mutiple(audio_track.display_name);
+    } else {
+      track_name = audio_track.display_name;
+    }
+  } else if (has_drc_streams) {
+    track_name = first_format.is_drc ? drc_labels.label_drc : drc_labels.label_original;
+  }
+
   const set: AudioSet = {
     mime_type: first_format.mime_type.split(';')[0],
     language: first_format.language ?? undefined,
     codecs: hoistCodecsIfPossible(formats, hoisted),
     audio_sample_rate: hoistNumberAttributeIfPossible(formats, 'audio_sample_rate', hoisted),
-    track_name: audio_track?.display_name,
-    track_role: getTrackRole(first_format),
+    track_name,
+    track_roles: getTrackRoles(first_format, has_drc_streams),
     channels: hoistAudioChannelsIfPossible(formats, hoisted),
     representations: formats.map((format) => getAudioRepresentation(format, hoisted, url_transformer, actions, player, cpn, shared_post_live_dvr_info))
   };
@@ -706,7 +743,8 @@ export function getStreamingInfo(
   cpn?: string,
   player?: Player,
   actions?: Actions,
-  storyboards?: PlayerStoryboardSpec | PlayerLiveStoryboardSpec
+  storyboards?: PlayerStoryboardSpec | PlayerLiveStoryboardSpec,
+  options?: StreamingInfoOptions
 ) {
   if (!streaming_data)
     throw new InnertubeError('Streaming data not available');
@@ -768,7 +806,17 @@ export function getStreamingInfo(
     audio_groups: [] as Format[][]
   });
 
-  const audio_sets = audio_groups.map((formats) => getAudioSet(formats, url_transformer, actions, player, cpn, shared_post_live_dvr_info));
+  let drc_labels: DrcLabels | undefined;
+
+  if (audio_groups.flat().some((format) => format.is_drc)) {
+    drc_labels = {
+      label_original: options?.label_original || 'Original',
+      label_drc: options?.label_drc || 'Stable Volume',
+      label_drc_mutiple: options?.label_drc_mutiple || ((display_name) => `${display_name} (Stable Volume)`)
+    };
+  }
+
+  const audio_sets = audio_groups.map((formats) => getAudioSet(formats, url_transformer, actions, player, cpn, shared_post_live_dvr_info, drc_labels));
 
   const video_sets = video_groups.map((formats) => getVideoSet(formats, url_transformer, player, actions, cpn, shared_post_live_dvr_info));