Identify audio stream languages (#847)

Tyrrrz · Dec 2, 2024 · 5062dc3 · 5062dc3
1 parent f8183ec
commit 5062dc3
Show file tree

Hide file tree

Showing 13 changed files with 214 additions and 9 deletions.
diff --git a/YoutubeExplode.Converter.Tests/GeneralSpecs.cs b/YoutubeExplode.Converter.Tests/GeneralSpecs.cs
@@ -92,7 +92,7 @@ public async Task I_can_download_a_video_as_a_single_mp4_file_with_multiple_stre
         var filePath = Path.Combine(dir.Path, "video.mp4");
 
         // Act
-        var manifest = await youtube.Videos.Streams.GetManifestAsync("9bZkp7q19f0");
+        var manifest = await youtube.Videos.Streams.GetManifestAsync("ngqcjXfggHQ");
 
         var audioStreamInfos = manifest
             .GetAudioOnlyStreams()
@@ -117,6 +117,20 @@ await youtube.Videos.DownloadAsync(
         // Assert
         MediaFormat.IsMp4File(filePath).Should().BeTrue();
 
+        foreach (var streamInfo in audioStreamInfos)
+        {
+            if (streamInfo.AudioLanguage is not null)
+            {
+                FileEx
+                    .ContainsBytes(
+                        filePath,
+                        Encoding.ASCII.GetBytes(streamInfo.AudioLanguage.Value.Name)
+                    )
+                    .Should()
+                    .BeTrue();
+            }
+        }
+
         foreach (var streamInfo in videoStreamInfos)
         {
             FileEx
@@ -136,7 +150,7 @@ public async Task I_can_download_a_video_as_a_single_webm_file_with_multiple_str
         var filePath = Path.Combine(dir.Path, "video.webm");
 
         // Act
-        var manifest = await youtube.Videos.Streams.GetManifestAsync("9bZkp7q19f0");
+        var manifest = await youtube.Videos.Streams.GetManifestAsync("ngqcjXfggHQ");
 
         var audioStreamInfos = manifest
             .GetAudioOnlyStreams()
@@ -161,6 +175,20 @@ await youtube.Videos.DownloadAsync(
         // Assert
         MediaFormat.IsWebMFile(filePath).Should().BeTrue();
 
+        foreach (var streamInfo in audioStreamInfos)
+        {
+            if (streamInfo.AudioLanguage is not null)
+            {
+                FileEx
+                    .ContainsBytes(
+                        filePath,
+                        Encoding.ASCII.GetBytes(streamInfo.AudioLanguage.Value.Name)
+                    )
+                    .Should()
+                    .BeTrue();
+            }
+        }
+
         foreach (var streamInfo in videoStreamInfos)
         {
             FileEx

diff --git a/YoutubeExplode.Converter/Converter.cs b/YoutubeExplode.Converter/Converter.cs
@@ -116,18 +116,43 @@ private async ValueTask ProcessAsync(
 
                 if (streamInput.Info is IAudioStreamInfo audioStreamInfo)
                 {
-                    arguments
-                        .Add($"-metadata:s:a:{lastAudioStreamIndex++}")
-                        .Add($"title={audioStreamInfo.Bitrate}");
+                    // Contains language information
+                    if (audioStreamInfo.AudioLanguage is not null)
+                    {
+                        // Language codes can be stored in any format, but most players expect
+                        // three-letter codes, so we'll try to convert to that first.
+                        var languageCode =
+                            audioStreamInfo.AudioLanguage.Value.TryGetThreeLetterCode()
+                            ?? audioStreamInfo.AudioLanguage.Value.Code;
+
+                        arguments
+                            .Add($"-metadata:s:a:{lastAudioStreamIndex}")
+                            .Add($"language={languageCode}")
+                            .Add($"-metadata:s:a:{lastAudioStreamIndex}")
+                            .Add(
+                                $"title={audioStreamInfo.AudioLanguage.Value.Name} | {audioStreamInfo.Bitrate}"
+                            );
+                    }
+                    // Does not contain language information
+                    else
+                    {
+                        arguments
+                            .Add($"-metadata:s:a:{lastAudioStreamIndex}")
+                            .Add($"title={audioStreamInfo.Bitrate}");
+                    }
+
+                    lastAudioStreamIndex++;
                 }
 
                 if (streamInput.Info is IVideoStreamInfo videoStreamInfo)
                 {
                     arguments
-                        .Add($"-metadata:s:v:{lastVideoStreamIndex++}")
+                        .Add($"-metadata:s:v:{lastVideoStreamIndex}")
                         .Add(
                             $"title={videoStreamInfo.VideoQuality.Label} | {videoStreamInfo.Bitrate}"
                         );
+
+                    lastVideoStreamIndex++;
                 }
             }
         }

diff --git a/YoutubeExplode.Tests/StreamSpecs.cs b/YoutubeExplode.Tests/StreamSpecs.cs
@@ -1,3 +1,4 @@
+using System;
 using System.Buffers;
 using System.IO;
 using System.Linq;
@@ -57,6 +58,61 @@ public async Task I_can_get_the_list_of_available_streams_of_a_video()
             .Contain(s => s.VideoQuality.MaxHeight == 144 && !s.VideoQuality.IsHighDefinition);
     }
 
+    [Fact]
+    public async Task I_can_get_the_list_of_available_streams_of_a_video_with_multiple_audio_languages()
+    {
+        // Arrange
+        var youtube = new YoutubeClient();
+
+        // Act
+        var manifest = await youtube.Videos.Streams.GetManifestAsync(
+            VideoIds.WithMultipleAudioLanguages
+        );
+
+        // Assert
+        manifest.Streams.Should().NotBeEmpty();
+
+        manifest
+            .GetAudioStreams()
+            .Should()
+            .Contain(t =>
+                t.AudioLanguage != null
+                && t.AudioLanguage.Value.Code == "en-US"
+                && t.AudioLanguage.Value.Name == "English (United States) original"
+                && t.IsAudioLanguageDefault == true
+            );
+
+        manifest
+            .GetAudioStreams()
+            .Should()
+            .Contain(t =>
+                t.AudioLanguage != null
+                && t.AudioLanguage.Value.Code == "fr-FR"
+                && t.AudioLanguage.Value.Name == "French (France)"
+                && t.IsAudioLanguageDefault == false
+            );
+
+        manifest
+            .GetAudioStreams()
+            .Should()
+            .Contain(t =>
+                t.AudioLanguage != null
+                && t.AudioLanguage.Value.Code == "it"
+                && t.AudioLanguage.Value.Name == "Italian"
+                && t.IsAudioLanguageDefault == false
+            );
+
+        manifest
+            .GetAudioStreams()
+            .Should()
+            .Contain(t =>
+                t.AudioLanguage != null
+                && t.AudioLanguage.Value.Code == "pt-BR"
+                && t.AudioLanguage.Value.Name == "Portuguese (Brazil)"
+                && t.IsAudioLanguageDefault == false
+            );
+    }
+
     [Theory]
     [InlineData(VideoIds.Normal)]
     [InlineData(VideoIds.Unlisted)]

diff --git a/YoutubeExplode.Tests/TestData/VideoIds.cs b/YoutubeExplode.Tests/TestData/VideoIds.cs
@@ -21,4 +21,5 @@ internal static class VideoIds
     public const string WithHighDynamicRangeStreams = "vX2vsvdq8nw";
     public const string WithClosedCaptions = "YltHGKX80Y8";
     public const string WithBrokenClosedCaptions = "1VKIIw05JnE";
+    public const string WithMultipleAudioLanguages = "ngqcjXfggHQ";
 }
diff --git a/YoutubeExplode/Bridge/DashManifest.cs b/YoutubeExplode/Bridge/DashManifest.cs
@@ -70,6 +70,12 @@ public class StreamData(XElement content) : IStreamData
         [Lazy]
         public string? AudioCodec => IsAudioOnly ? (string?)content.Attribute("codecs") : null;
 
+        public string? AudioLanguageCode => null;
+
+        public string? AudioLanguageName => null;
+
+        public bool? IsAudioLanguageDefault => null;
+
         [Lazy]
         public string? VideoCodec => IsAudioOnly ? null : (string?)content.Attribute("codecs");
 

diff --git a/YoutubeExplode/Bridge/IStreamData.cs b/YoutubeExplode/Bridge/IStreamData.cs
@@ -18,6 +18,12 @@ internal interface IStreamData
 
     string? AudioCodec { get; }
 
+    string? AudioLanguageCode { get; }
+
+    string? AudioLanguageName { get; }
+
+    bool? IsAudioLanguageDefault { get; }
+
     string? VideoCodec { get; }
 
     string? VideoQualityLabel { get; }

diff --git a/YoutubeExplode/Bridge/PlayerResponse.cs b/YoutubeExplode/Bridge/PlayerResponse.cs
@@ -246,6 +246,28 @@ public class StreamData(JsonElement content) : IStreamData
         public string? AudioCodec =>
             IsAudioOnly ? Codecs : Codecs?.SubstringAfter(", ").NullIfWhiteSpace();
 
+        [Lazy]
+        public string? AudioLanguageCode =>
+            content
+                .GetPropertyOrNull("audioTrack")
+                ?.GetPropertyOrNull("id")
+                ?.GetStringOrNull()
+                ?.SubstringUntil(".");
+
+        [Lazy]
+        public string? AudioLanguageName =>
+            content
+                .GetPropertyOrNull("audioTrack")
+                ?.GetPropertyOrNull("displayName")
+                ?.GetStringOrNull();
+
+        [Lazy]
+        public bool? IsAudioLanguageDefault =>
+            content
+                .GetPropertyOrNull("audioTrack")
+                ?.GetPropertyOrNull("audioIsDefault")
+                ?.GetBooleanOrNull();
+
         [Lazy]
         public string? VideoCodec
         {

diff --git a/...Explode/Videos/ClosedCaptions/Language.cs → YoutubeExplode/Common/Language.cs b/...Explode/Videos/ClosedCaptions/Language.cs → YoutubeExplode/Common/Language.cs
@@ -1,6 +1,8 @@
 using System;
 using System.Diagnostics.CodeAnalysis;
 
+// TODO: breaking change: update the namespace
+// ReSharper disable once CheckNamespace
 namespace YoutubeExplode.Videos.ClosedCaptions;
 
 /// <summary>

diff --git a/YoutubeExplode/Utils/Extensions/JsonExtensions.cs b/YoutubeExplode/Utils/Extensions/JsonExtensions.cs
@@ -25,6 +25,14 @@ internal static class JsonExtensions
         return null;
     }
 
+    public static bool? GetBooleanOrNull(this JsonElement element) =>
+        element.ValueKind switch
+        {
+            JsonValueKind.True => true,
+            JsonValueKind.False => false,
+            _ => null,
+        };
+
     public static string? GetStringOrNull(this JsonElement element) =>
         element.ValueKind == JsonValueKind.String ? element.GetString() : null;
 

diff --git a/YoutubeExplode/Videos/Streams/AudioOnlyStreamInfo.cs b/YoutubeExplode/Videos/Streams/AudioOnlyStreamInfo.cs
@@ -1,4 +1,5 @@
 using System.Diagnostics.CodeAnalysis;
+using YoutubeExplode.Videos.ClosedCaptions;
 
 namespace YoutubeExplode.Videos.Streams;
 
@@ -10,7 +11,9 @@ public class AudioOnlyStreamInfo(
     Container container,
     FileSize size,
     Bitrate bitrate,
-    string audioCodec
+    string audioCodec,
+    Language? audioLanguage,
+    bool? isAudioLanguageDefault
 ) : IAudioStreamInfo
 {
     /// <inheritdoc />
@@ -28,7 +31,16 @@ string audioCodec
     /// <inheritdoc />
     public string AudioCodec { get; } = audioCodec;
 
+    /// <inheritdoc />
+    public Language? AudioLanguage { get; } = audioLanguage;
+
+    /// <inheritdoc />
+    public bool? IsAudioLanguageDefault { get; } = isAudioLanguageDefault;
+
     /// <inheritdoc />
     [ExcludeFromCodeCoverage]
-    public override string ToString() => $"Audio-only ({Container})";
+    public override string ToString() =>
+        AudioLanguage is not null
+            ? $"Audio-only ({Container} | {AudioLanguage})"
+            : $"Audio-only ({Container})";
 }
diff --git a/YoutubeExplode/Videos/Streams/IAudioStreamInfo.cs b/YoutubeExplode/Videos/Streams/IAudioStreamInfo.cs
@@ -1,3 +1,5 @@
+using YoutubeExplode.Videos.ClosedCaptions;
+
 namespace YoutubeExplode.Videos.Streams;
 
 /// <summary>
@@ -9,4 +11,20 @@ public interface IAudioStreamInfo : IStreamInfo
     /// Audio codec.
     /// </summary>
     string AudioCodec { get; }
+
+    /// <summary>
+    /// Audio language.
+    /// </summary>
+    /// <remarks>
+    /// May be null if the audio stream does not contain language information.
+    /// </remarks>
+    Language? AudioLanguage { get; }
+
+    /// <summary>
+    /// Whether the audio stream's language corresponds to the default language of the video.
+    /// </summary>
+    /// <remarks>
+    /// May be null if the audio stream does not contain language information.
+    /// </remarks>
+    bool? IsAudioLanguageDefault { get; }
 }
diff --git a/YoutubeExplode/Videos/Streams/MuxedStreamInfo.cs b/YoutubeExplode/Videos/Streams/MuxedStreamInfo.cs
@@ -1,5 +1,6 @@
 using System.Diagnostics.CodeAnalysis;
 using YoutubeExplode.Common;
+using YoutubeExplode.Videos.ClosedCaptions;
 
 namespace YoutubeExplode.Videos.Streams;
 
@@ -12,6 +13,8 @@ public class MuxedStreamInfo(
     FileSize size,
     Bitrate bitrate,
     string audioCodec,
+    Language? audioLanguage,
+    bool? isAudioLanguageDefault,
     string videoCodec,
     VideoQuality videoQuality,
     Resolution videoResolution
@@ -32,6 +35,12 @@ Resolution videoResolution
     /// <inheritdoc />
     public string AudioCodec { get; } = audioCodec;
 
+    /// <inheritdoc />
+    public Language? AudioLanguage { get; } = audioLanguage;
+
+    /// <inheritdoc />
+    public bool? IsAudioLanguageDefault { get; } = isAudioLanguageDefault;
+
     /// <inheritdoc />
     public string VideoCodec { get; } = videoCodec;
 

diff --git a/YoutubeExplode/Videos/Streams/StreamClient.cs b/YoutubeExplode/Videos/Streams/StreamClient.cs
@@ -13,6 +13,7 @@
 using YoutubeExplode.Exceptions;
 using YoutubeExplode.Utils;
 using YoutubeExplode.Utils.Extensions;
+using YoutubeExplode.Videos.ClosedCaptions;
 
 namespace YoutubeExplode.Videos.Streams;
 
@@ -123,6 +124,13 @@ private async IAsyncEnumerable<IStreamInfo> GetStreamInfosAsync(
                 streamData.Bitrate?.Pipe(s => new Bitrate(s))
                 ?? throw new YoutubeExplodeException("Failed to extract the stream bitrate.");
 
+            var audioLanguage = !string.IsNullOrWhiteSpace(streamData.AudioLanguageCode)
+                ? new Language(
+                    streamData.AudioLanguageCode,
+                    streamData.AudioLanguageName ?? streamData.AudioLanguageCode
+                )
+                : (Language?)null;
+
             // Muxed or video-only stream
             if (!string.IsNullOrWhiteSpace(streamData.VideoCodec))
             {
@@ -146,6 +154,8 @@ streamData.VideoWidth is not null && streamData.VideoHeight is not null
                         new FileSize(contentLength.Value),
                         bitrate,
                         streamData.AudioCodec,
+                        audioLanguage,
+                        streamData.IsAudioLanguageDefault,
                         streamData.VideoCodec,
                         videoQuality,
                         videoResolution
@@ -177,7 +187,9 @@ streamData.VideoWidth is not null && streamData.VideoHeight is not null
                     container,
                     new FileSize(contentLength.Value),
                     bitrate,
-                    streamData.AudioCodec
+                    streamData.AudioCodec,
+                    audioLanguage,
+                    streamData.IsAudioLanguageDefault
                 );
 
                 yield return streamInfo;