Add more metadata fields to OGG and MP4 downloads

TobiGr · TobiGr · commit 51741e38c997 · 2026-02-08T13:26:24.000+01:00
Copyright, record label, album, label, ...
diff --git a/app/src/main/java/org/schabi/newpipe/streams/Mp4FromDashWriter.java b/app/src/main/java/org/schabi/newpipe/streams/Mp4FromDashWriter.java
@@ -11,6 +11,7 @@
 import org.schabi.newpipe.streams.Mp4DashReader.TrackKind;
 import org.schabi.newpipe.streams.Mp4DashReader.TrunEntry;
 import org.schabi.newpipe.streams.io.SharpStream;
+import org.schabi.newpipe.util.StreamInfoMetadataHelper;
 
 import java.io.ByteArrayOutputStream;
 import java.io.IOException;
@@ -930,17 +931,21 @@ private byte[] makeSgpd() {
 
     /**
      * Create the 'udta' box with metadata fields.
+     * {@code udta} is a user data box that can contain various types of metadata,
+     * including title, artist, date, and cover art.
+     * @see <a href="https://developer.apple.com/documentation/quicktime-file-format/
+     * user_data_atoms">Apple Quick Time Format Specification for user data atoms</a>
+     * @see <a href="https://wiki.multimedia.cx/index.php?title=FFmpeg_Metadata
+     * #QuickTime/MOV/MP4/M4A/et_al.">Multimedia Wiki FFmpeg Metadata</a>
+     * @see <a href="https://atomicparsley.sourceforge.net/mpeg-4files.html">atomicparsley docs</a>
+     * for a short and understandable reference about metadata keys and values
      * @throws IOException
      */
     private void makeUdta() throws IOException {
         if (streamInfo == null) {
             return;
         }
 
-        final String title = streamInfo.getName();
-        final String artist = streamInfo.getUploaderName();
-        final String date = streamInfo.getUploadDate().getLocalDateTime().toLocalDate().toString();
-
         // udta
         final int startUdta = auxOffset();
         auxWrite(ByteBuffer.allocate(8).putInt(0).putInt(0x75647461).array()); // "udta"
@@ -957,6 +962,16 @@ private void makeUdta() throws IOException {
         final int startIlst = auxOffset();
         auxWrite(ByteBuffer.allocate(8).putInt(0).putInt(0x696C7374).array()); // "ilst"
 
+        // write metadata items
+
+        final var metaHelper = new StreamInfoMetadataHelper(streamInfo);
+        final String title = metaHelper.getTitle();
+        final String artist = metaHelper.getArtist();
+        final String date = metaHelper.getReleaseDate().getLocalDateTime()
+                .toLocalDate().toString();
+        final String recordLabel = metaHelper.getRecordLabel();
+        final String copyright = metaHelper.getCopyright();
+
         if (title != null && !title.isEmpty()) {
             writeMetaItem("©nam", title);
         }
@@ -967,8 +982,12 @@ private void makeUdta() throws IOException {
             // this means 'year' in mp4 metadata, who the hell thought that?
             writeMetaItem("©day", date);
         }
-
-
+        if (recordLabel != null && !recordLabel.isEmpty()) {
+            writeMetaItem("©lab", recordLabel);
+        }
+        if (copyright != null && !copyright.isEmpty()) {
+            writeMetaItem("©cpy", copyright);
+        }
 
         if (thumbnail != null) {
             final ByteArrayOutputStream baos = new ByteArrayOutputStream();
diff --git a/app/src/main/java/org/schabi/newpipe/streams/OggFromWebMWriter.java b/app/src/main/java/org/schabi/newpipe/streams/OggFromWebMWriter.java
@@ -9,12 +9,14 @@
 import androidx.annotation.NonNull;
 import androidx.annotation.Nullable;
 
+import org.schabi.newpipe.extractor.stream.SongMetadata;
 import org.schabi.newpipe.extractor.stream.StreamInfo;
 import org.schabi.newpipe.streams.WebMReader.Cluster;
 import org.schabi.newpipe.streams.WebMReader.Segment;
 import org.schabi.newpipe.streams.WebMReader.SimpleBlock;
 import org.schabi.newpipe.streams.WebMReader.WebMTrack;
 import org.schabi.newpipe.streams.io.SharpStream;
+import org.schabi.newpipe.util.StreamInfoMetadataHelper;
 
 import java.io.ByteArrayOutputStream;
 import java.io.Closeable;
@@ -40,12 +42,18 @@
  * </p>
  * <ul>
  *     <li>FLAC: <a href="https://www.rfc-editor.org/rfc/rfc9639">RFC 9639</a></li>
+ *     <li>
+ *         Vorbis: <a href="https://www.xiph.org/vorbis/doc/Vorbis_I_spec.html">Vorbis I</a>.
+ *         <br>
+ *         Vorbis uses FLAC picture blocks for embedding cover art in the metadata.
+ *     </li>
  *     <li>Opus: All specs can be found at <a href="https://opus-codec.org/docs/">
  *         https://opus-codec.org/docs/</a>.
  *         <a href="https://datatracker.ietf.org/doc/html/rfc7845.html">RFC7845</a>
  *         defines the Ogg encapsulation for Opus streams, i.e.the container format and metadata.
+ *         <br>
+ *         Opus uses multiple Vorbis I features, e.g. the comment header format for metadata.
  *     </li>
- *     <li>Vorbis: <a href="https://www.xiph.org/vorbis/doc/Vorbis_I_spec.html">Vorbis I</a></li>
  * </ul>
  *
  * @author kapodamy
@@ -352,8 +360,8 @@ private int makePacketHeader(final long granPos, @NonNull final ByteBuffer buffe
      * @see <a href="https://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-610004.2">
      *     Vorbis I 4.2. Header decode and decode setup</a> and
      *     <a href="https://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-820005">
-     *     Vorbis 5. comment field and header specification</a>
-     *     for VORBIS metadata header format
+     *     Vorbis I 5. comment field and header specification</a>
+     *     for VORBIS metadata header format. Vorbis I 5. lists all the possible metadata tags.
      *
      * @return the metadata header as a byte array, or null if the codec is not supported
      * for metadata generation
@@ -364,38 +372,58 @@ private byte[] makeCommentHeader() {
             Log.d(TAG, "Downloading media with codec ID " + webmTrack.codecId);
         }
 
-        if ("A_OPUS".equals(webmTrack.codecId)) {
-            final var metadata = new ArrayList<Pair<String, String>>();
-            if (streamInfo != null) {
-                metadata.add(Pair.create("COMMENT", streamInfo.getUrl()));
-                metadata.add(Pair.create("GENRE", streamInfo.getCategory()));
-                metadata.add(Pair.create("ARTIST", streamInfo.getUploaderName()));
-                metadata.add(Pair.create("TITLE", streamInfo.getName()));
-                metadata.add(Pair.create("DATE", streamInfo
-                        .getUploadDate()
-                        .getLocalDateTime()
-                        .format(DateTimeFormatter.ISO_DATE)));
-                 if (thumbnail != null) {
-                     metadata.add(makeFlacPictureTag(thumbnail));
-                 }
+        final var metadata = new ArrayList<Pair<String, String>>();
+        if (streamInfo != null) {
+            final SongMetadata songMetadata = streamInfo.getSongMetadata();
+            final StreamInfoMetadataHelper metadHelper = new StreamInfoMetadataHelper(streamInfo);
+            // metadata that can be present in the stream info and the song metadata.
+            // Use the song metadata if available, otherwise fallback to stream info.
+            metadata.add(Pair.create("COMMENT", streamInfo.getUrl()));
+            metadata.add(Pair.create("GENRE", metadHelper.getGenre()));
+            metadata.add(Pair.create("ARTIST", metadHelper.getArtist()));
+            metadata.add(Pair.create("TITLE", metadHelper.getTitle()));
+            metadata.add(Pair.create("DATE", metadHelper.getReleaseDate()
+                    .getLocalDateTime()
+                    .format(DateTimeFormatter.ISO_DATE)));
+            // Additional metadata that is only present in the song metadata
+            if (songMetadata != null) {
+                metadata.add(Pair.create("ALBUM", songMetadata.album));
+                if (songMetadata.track != SongMetadata.TRACK_UNKNOWN) {
+                    // TRACKNUMBER is suggested in Vorbis spec,
+                    // but TRACK is more commonly used in practice
+                    metadata.add(Pair.create("TRACKNUMBER", String.valueOf(songMetadata.track)));
+                    metadata.add(Pair.create("TRACK", String.valueOf(songMetadata.track)));
+                }
+                metadata.add(Pair.create("PERFORMER", String.join(", ", songMetadata.performer)));
+                metadata.add(Pair.create("ORGANIZATION", songMetadata.label));
+                metadata.add(Pair.create("COPYRIGHT", songMetadata.copyright));
             }
-
-            if (DEBUG) {
-                Log.d(TAG, "Creating metadata header with this data:");
-                metadata.forEach(p -> Log.d(TAG, p.first + "=" + p.second));
+            // Add thumbnail as cover art at the end because it is the largest metadata entry
+            if (thumbnail != null) {
+                metadata.add(makeFlacPictureTag(thumbnail));
             }
+        }
+
+        if (DEBUG) {
+            Log.d(TAG, "Creating metadata header with this data:");
+            metadata.forEach(p -> Log.d(TAG, p.first + "=" + p.second));
+        }
 
-            return makeOpusTagsHeader(metadata);
+        if ("A_OPUS".equals(webmTrack.codecId)) {
+            // See RFC7845 5.2: https://datatracker.ietf.org/doc/html/rfc7845.html#section-5.2
+            final byte[] identificationHeader = new byte[]{
+                    0x4F, 0x70, 0x75, 0x73, 0x54, 0x61, 0x67, 0x73, // "OpusTags" binary string
+                    0x00, 0x00, 0x00, 0x00, // vendor (aka. Encoder) string of length 0
+            };
+            return makeCommentHeader(metadata, identificationHeader);
         } else if ("A_VORBIS".equals(webmTrack.codecId)) {
             // See https://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-610004.2
-            // for the Vorbis comment header format
-            // TODO: add Vorbis metadata: same as Opus, but with the Vorbis comment header format
-            return new byte[]{
+            final byte[] identificationHeader = new byte[]{
                     0x03, // packet type for Vorbis comment header
                     0x76, 0x6f, 0x72, 0x62, 0x69, 0x73, // "vorbis" binary string
-                    0x00, 0x00, 0x00, 0x00, // writing application string size (not present)
-                    0x00, 0x00, 0x00, 0x00 // additional tags count (zero means no tags)
+                    0x00, 0x00, 0x00, 0x00, // vendor (aka. Encoder) string of length 0
             };
+            return makeCommentHeader(metadata, identificationHeader);
         }
 
         // not implemented for the desired codec
@@ -405,12 +433,12 @@ private byte[] makeCommentHeader() {
     /**
      * This creates a single metadata tag for use in opus metadata headers. It contains the four
      * byte string length field and includes the string as-is. This cannot be used independently,
-     * but must follow a proper "OpusTags" header.
+     * but must follow a proper Comment header.
      *
      * @param pair A key-value pair in the format "KEY=some value"
      * @return The binary data of the encoded metadata tag
      */
-    private static byte[] makeOpusMetadataTag(final Pair<String, String> pair) {
+    private static byte[] makeVorbisMetadataTag(final Pair<String, String> pair) {
         final var keyValue = pair.first.toUpperCase() + "=" + pair.second.trim();
 
         final var bytes = keyValue.getBytes();
@@ -486,24 +514,21 @@ private static Pair<String, String> makeFlacPictureTag(final Bitmap bitmap) {
     }
 
     /**
-     * This returns a complete "OpusTags" header, created from the provided metadata tags.
-     * <p>
-     * You probably want to use makeOpusMetadata(), which uses this function to create
-     * a header with sensible metadata filled in.
-     *
-     * @ImplNote See <a href="https://datatracker.ietf.org/doc/html/rfc7845.html#section-5.2">
-     *     RFC7845 5.2</a>
+     * This returns a complete Comment header, created from the provided metadata tags.
      *
      * @param keyValueLines A list of pairs of the tags. This can also be though of as a mapping
      *                      from one key to multiple values.
+     * @param identificationHeader the identification header for the codec,
+     *                             which is required to be prefixed to the comment header.
      * @return The binary header
      */
-    private static byte[] makeOpusTagsHeader(final List<Pair<String, String>> keyValueLines) {
+    private static byte[] makeCommentHeader(final List<Pair<String, String>> keyValueLines,
+                                            final byte[] identificationHeader) {
         final var tags = keyValueLines
                 .stream()
-                .filter(p -> !p.second.isBlank())
-                .map(OggFromWebMWriter::makeOpusMetadataTag)
-                .collect(Collectors.toUnmodifiableList());
+                .filter(p -> p.second != null && !p.second.isBlank())
+                .map(OggFromWebMWriter::makeVorbisMetadataTag)
+                .toList();
 
         final var tagsBytes = tags.stream().collect(Collectors.summingInt(arr -> arr.length));
 
@@ -512,11 +537,7 @@ private static byte[] makeOpusTagsHeader(final List<Pair<String, String>> keyVal
 
         final var head = ByteBuffer.allocate(byteCount);
         head.order(ByteOrder.LITTLE_ENDIAN);
-        // See RFC7845 5.2: https://datatracker.ietf.org/doc/html/rfc7845.html#section-5.2
-        head.put(new byte[]{
-                0x4F, 0x70, 0x75, 0x73, 0x54, 0x61, 0x67, 0x73, // "OpusTags" binary string
-                0x00, 0x00, 0x00, 0x00, // vendor (aka. Encoder) string of length 0
-        });
+        head.put(identificationHeader);
         head.putInt(tags.size()); // 4 bytes for tag count
         tags.forEach(head::put); // dynamic amount of tag bytes
 
diff --git a/app/src/main/java/org/schabi/newpipe/util/StreamInfoMetadataHelper.kt b/app/src/main/java/org/schabi/newpipe/util/StreamInfoMetadataHelper.kt
@@ -0,0 +1,57 @@
+package org.schabi.newpipe.util
+
+import org.schabi.newpipe.extractor.localization.DateWrapper
+import org.schabi.newpipe.extractor.stream.SongMetadata
+import org.schabi.newpipe.extractor.stream.StreamInfo
+
+class StreamInfoMetadataHelper(
+    val streamInfo: StreamInfo
+) {
+    val songInfo: SongMetadata? = streamInfo.songMetadata
+
+    fun getTitle(): String? {
+        if (songInfo?.title?.contentEquals(streamInfo.name) == true) {
+            // YT Music uses uppercase chars in the description, but the StreamInfo name is using
+            // the correct case, so we prefer that
+            return streamInfo.name
+        }
+        return if (songInfo?.title?.isBlank() == false) songInfo.title else streamInfo.name
+    }
+
+    fun getArtist(): String? {
+        if (songInfo?.artist?.contentEquals(streamInfo.uploaderName) == true) {
+            // YT Music uses uppercase chars in the description, but the uploader name is using
+            // the correct case, so we prefer the uploader name
+            return streamInfo.uploaderName
+        }
+        return if (songInfo?.artist?.isBlank() == false) {
+            songInfo.artist
+        } else {
+            streamInfo.uploaderName
+        }
+    }
+
+    fun getPerformer(): List<String?>? = songInfo?.performer
+
+    fun getComposer(): String? = songInfo?.composer
+
+    fun getGenre(): String? = if (songInfo?.genre?.isEmpty() == false) {
+        songInfo.genre
+    } else {
+        streamInfo.category
+    }
+
+    fun getAlbum(): String? = songInfo?.album
+
+    fun getTrackNumber(): Int? = songInfo?.track
+
+    fun getDuration(): Long = songInfo?.duration?.seconds ?: streamInfo.duration
+
+    fun getReleaseDate(): DateWrapper = songInfo?.releaseDate ?: streamInfo.uploadDate
+
+    fun getRecordLabel(): String? = songInfo?.label
+
+    fun getCopyright(): String? = songInfo?.copyright ?: streamInfo.licence
+
+    fun getLocation(): String? = songInfo?.location
+}
