Merge pull request #874 from n3d1117/feat/mpv-audio-track-switching

feat: MPV audio track switching and fix default audio language
2025-10-30 20:22:06 +00:00 · 2025-06-17 16:02:09 +02:00
parent 2a597ab3cb 2461a33feb
commit e41527775a
7 changed files with 216 additions and 12 deletions
--- a/Model/Applications/InvidiousAPI.swift
+++ b/Model/Applications/InvidiousAPI.swift
@@ -685,50 +685,93 @@ final class InvidiousAPI: Service, ObservableObject, VideosAPI {
        }
    }

+    func extractXTags(from urlString: String) -> [String: String] {
+        guard let urlComponents = URLComponents(string: urlString),
+              let queryItems = urlComponents.queryItems,
+              let xtagsValue = queryItems.first(where: { $0.name == "xtags" })?.value else {
+            return [:]
+        }
+
+        guard let decoded = xtagsValue.removingPercentEncoding else { return [:] }
+
+        // Parse key-value pairs (format: key1=value1:key2=value2)
+        // Example: "acont=dubbed-auto:lang=en-US"
+        let pairs = decoded.split(separator: ":")
+        var result: [String: String] = [:]
+        for pair in pairs {
+            let parts = pair.split(separator: "=", maxSplits: 1)
+            if parts.count == 2 {
+                result[String(parts[0])] = String(parts[1])
+            }
+        }
+        return result
+    }
+
    private func extractAdaptiveFormats(from streams: [JSON], videoId: String?) -> [Stream] {
-        let audioStreams = streams
+        let audioTracks = streams
            .filter { $0["type"].stringValue.starts(with: "audio/mp4") }
            .sorted {
                $0.dictionaryValue["bitrate"]?.int ?? 0 >
                    $1.dictionaryValue["bitrate"]?.int ?? 0
            }
-        guard let audioStream = audioStreams.first else {
+            .compactMap { audioStream -> Stream.AudioTrack? in
+                guard let url = audioStream["url"].url,
+                      let audioItag = audioStream["itag"].string
+                else { return nil }
+
+                let finalURL: URL
+                if let videoId, account.instance.invidiousCompanion {
+                    let audioCompanionURLString = "\(account.instance.apiURLString)/latest_version?id=\(videoId)&itag=\(audioItag)"
+                    finalURL = URL(string: audioCompanionURLString) ?? url
+                } else {
+                    finalURL = url
+                }
+
+                let xTags = extractXTags(from: url.absoluteString)
+
+                return Stream.AudioTrack(
+                    url: finalURL,
+                    content: xTags["acont"],
+                    language: xTags["lang"]
+                )
+            }
+            .sorted {
+                /// Always prefer original audio streams over dubbed ones
+                !$0.isDubbed && $1.isDubbed
+            }
+
+        guard !audioTracks.isEmpty else {
            return .init()
        }

        let videoStreams = streams.filter { $0["type"].stringValue.starts(with: "video/") }

        return videoStreams.compactMap { videoStream in
-            guard let audioAssetURL = audioStream["url"].url,
-                  let videoAssetURL = videoStream["url"].url,
-                  let audioItag = audioStream["itag"].string,
+            guard let videoAssetURL = videoStream["url"].url,
                  let videoItag = videoStream["itag"].string
            else {
                return nil
            }
-            let finalAudioURL: URL
            let finalVideoURL: URL

            if let videoId, account.instance.invidiousCompanion {
-                let audioCompanionURLString = "\(account.instance.apiURLString)/latest_version?id=\(videoId)&itag=\(audioItag)"
                let videoCompanionURLString = "\(account.instance.apiURLString)/latest_version?id=\(videoId)&itag=\(videoItag)"
-                finalAudioURL = URL(string: audioCompanionURLString) ?? audioAssetURL
                finalVideoURL = URL(string: videoCompanionURLString) ?? videoAssetURL
            } else {
-                finalAudioURL = audioAssetURL
                finalVideoURL = videoAssetURL
            }

            return Stream(
                instance: account.instance,
-                audioAsset: AVURLAsset(url: finalAudioURL),
+                audioAsset: AVURLAsset(url: audioTracks[0].url),
                videoAsset: AVURLAsset(url: finalVideoURL),
                resolution: Stream.Resolution.from(resolution: videoStream["resolution"].stringValue),
                kind: .adaptive,
                encoding: videoStream["encoding"].string,
                videoFormat: videoStream["type"].string,
                bitrate: videoStream["bitrate"].int,
-                requestRange: videoStream["init"].string ?? videoStream["index"].string
+                requestRange: videoStream["init"].string ?? videoStream["index"].string,
+                audioTracks: audioTracks
            )
        }
    }
--- a/Model/Player/Backends/MPVBackend.swift
+++ b/Model/Player/Backends/MPVBackend.swift
@@ -185,6 +185,10 @@ final class MPVBackend: PlayerBackend {
    var audioSampleRate: String {
        client?.audioSampleRate ?? "unknown"
    }
+    
+    var availableAudioTracks: [Stream.AudioTrack] {
+        stream?.audioTracks ?? []
+    }

    init() {
        clientTimer = .init(interval: .seconds(Self.timeUpdateInterval), mode: .infinite) { [weak self] _ in
@@ -243,6 +247,9 @@ final class MPVBackend: PlayerBackend {

        let updateCurrentStream = {
            DispatchQueue.main.async { [weak self] in
+                if self?.video?.id != video.id {
+                    self?.model.selectedAudioTrackIndex = 0
+                }
                self?.stream = stream
                self?.video = video
                self?.model.stream = stream
@@ -319,6 +326,7 @@ final class MPVBackend: PlayerBackend {
                        startPlaying()
                    }

+                    stream.audioAsset = AVURLAsset(url: stream.audioTracks[stream.selectedAudioTrackIndex].url)
                    let fileToLoad = self.model.musicMode ? stream.audioAsset.url : stream.videoAsset.url
                    let audioTrack = self.model.musicMode ? nil : stream.audioAsset.url

@@ -728,4 +736,13 @@ final class MPVBackend: PlayerBackend {
            logger.info("MPV backend received unhandled property: \(name)")
        }
    }
+
+    func switchAudioTrack(to index: Int) {
+        guard let stream, let video else { return }
+
+        stream.selectedAudioTrackIndex = index
+        model.saveTime { [weak self] in
+            self?.playStream(stream, of: video, preservingTime: true, upgrading: false)
+        }
+    }
 }
--- a/Model/Player/PlayerModel.swift
+++ b/Model/Player/PlayerModel.swift
@@ -210,6 +210,14 @@ final class PlayerModel: ObservableObject {
        var keyPressMonitor: Any?
    #endif

+    @Published var selectedAudioTrackIndex = 0 {
+        didSet {
+            if oldValue != selectedAudioTrackIndex {
+                handleAudioTrackChange()
+            }
+        }
+    }
+
    init() {
        #if os(iOS)
            isOrientationLocked = Defaults[.isOrientationLocked]
@@ -1467,4 +1475,12 @@ final class PlayerModel: ObservableObject {
            }
        }
    #endif
+
+    private func handleAudioTrackChange() {
+        (backend as? MPVBackend)?.switchAudioTrack(to: selectedAudioTrackIndex)
+    }
+
+    var availableAudioTracks: [Stream.AudioTrack] {
+        (backend as? MPVBackend)?.availableAudioTracks ?? []
+    }
 }
--- a/Model/Stream.swift
+++ b/Model/Stream.swift
@@ -191,6 +191,25 @@ class Stream: Equatable, Hashable, Identifiable {
            return .unknown
        }
    }
+    
+    struct AudioTrack: Hashable, Identifiable {
+        let id = UUID().uuidString
+        let url: URL
+        let content: String?
+        let language: String?
+        
+        var displayLanguage: String {
+            LanguageCodes(rawValue: language ?? "")?.description.capitalized ?? language ?? "Unknown"
+        }
+        
+        var description: String {
+            "\(displayLanguage) (\(content ?? "Unknown"))"
+        }
+        
+        var isDubbed: Bool {
+            content?.lowercased().starts(with: "dubbed") ?? false
+        }
+    }

    let id = UUID()

@@ -208,6 +227,8 @@ class Stream: Equatable, Hashable, Identifiable {
    var videoFormat: String?
    var bitrate: Int?
    var requestRange: String?
+    var audioTracks: [AudioTrack] = []
+    var selectedAudioTrackIndex = 0

    init(
        instance: Instance? = nil,
@@ -220,7 +241,8 @@ class Stream: Equatable, Hashable, Identifiable {
        encoding: String? = nil,
        videoFormat: String? = nil,
        bitrate: Int? = nil,
-        requestRange: String? = nil
+        requestRange: String? = nil,
+        audioTracks: [AudioTrack] = []
    ) {
        self.instance = instance
        self.audioAsset = audioAsset
@@ -233,6 +255,7 @@ class Stream: Equatable, Hashable, Identifiable {
        format = .from(videoFormat ?? "")
        self.bitrate = bitrate
        self.requestRange = requestRange
+        self.audioTracks = audioTracks
    }

    var isLocal: Bool {
--- a/Shared/LanguageCodes.swift
+++ b/Shared/LanguageCodes.swift
@@ -11,6 +11,7 @@ enum LanguageCodes: String, CaseIterable {
    case Greek = "el"
    case English = "en"
    case English_GB = "en-GB"
+    case English_US = "en-US"
    case Spanish = "es"
    case Persian = "fa"
    case Finnish = "fi"
@@ -76,6 +77,8 @@ enum LanguageCodes: String, CaseIterable {
            return "English"
        case .English_GB:
            return "English (United Kingdom)"
+        case .English_US:
+            return "English (United States)"
        case .Spanish:
            return "Spanish"
        case .Persian:
--- a/Shared/Player/Controls/ControlsOverlay.swift
+++ b/Shared/Player/Controls/ControlsOverlay.swift
@@ -19,6 +19,7 @@ struct ControlsOverlay: View {
            case increaseRate
            case decreaseRate
            case captions
+            case audioTrack
        }

        @FocusState private var focusedField: Field?
@@ -60,6 +61,15 @@ struct ControlsOverlay: View {
                    #endif
                }

+                if !player.availableAudioTracks.isEmpty {
+                    Section(header: controlsHeader("Audio Track".localized())) {
+                        audioTrackButton
+                        #if os(tvOS)
+                        .focused($focusedField, equals: .audioTrack)
+                        #endif
+                    }
+                }
+
                Section(header: controlsHeader("Stream & Player".localized())) {
                    qualityButton
                    #if os(tvOS)
@@ -438,6 +448,46 @@ struct ControlsOverlay: View {
            }
        )
    }
+
+    @ViewBuilder private var audioTrackButton: some View {
+        #if os(macOS)
+            audioTrackPicker
+                .labelsHidden()
+                .frame(maxWidth: 300)
+        #elseif os(iOS)
+            Menu {
+                audioTrackPicker
+            } label: {
+                Text(player.availableAudioTracks[player.selectedAudioTrackIndex].displayLanguage)
+                    .frame(maxWidth: 240, alignment: .trailing)
+            }
+            .transaction { t in t.animation = .none }
+            .buttonStyle(.plain)
+            .foregroundColor(.accentColor)
+            .frame(maxWidth: 240, alignment: .trailing)
+            .frame(height: 40)
+        #else
+            ControlsOverlayButton(focusedField: $focusedField, field: .audioTrack) {
+                Text(player.availableAudioTracks[player.selectedAudioTrackIndex].displayLanguage)
+                    .frame(maxWidth: 320)
+            }
+            .contextMenu {
+                ForEach(Array(player.availableAudioTracks.enumerated()), id: \.offset) { index, track in
+                    Button(track.description) { player.selectedAudioTrackIndex = index }
+                }
+                Button("Cancel", role: .cancel) {}
+            }
+        #endif
+    }
+
+    private var audioTrackPicker: some View {
+        Picker("", selection: $player.selectedAudioTrackIndex) {
+            ForEach(Array(player.availableAudioTracks.enumerated()), id: \.offset) { index, track in
+                Text(track.description).tag(index)
+            }
+        }
+        .transaction { t in t.animation = .none }
+    }
 }

 struct ControlsOverlay_Previews: PreviewProvider {
--- a/Shared/Player/PlaybackSettings.swift
+++ b/Shared/Player/PlaybackSettings.swift
@@ -22,6 +22,7 @@ struct PlaybackSettings: View {
            case increaseRate
            case decreaseRate
            case captions
+            case audioTrack
        }

        @FocusState private var focusedField: Field?
@@ -112,6 +113,17 @@ struct PlaybackSettings: View {
                    #endif
                }

+                if !player.availableAudioTracks.isEmpty {
+                    HStack {
+                        controlsHeader("Audio Track".localized())
+                        Spacer()
+                        audioTrackButton
+                        #if os(tvOS)
+                        .focused($focusedField, equals: .audioTrack)
+                        #endif
+                    }
+                }
+
                HStack(spacing: 8) {
                    controlsHeader("Backend".localized())
                    Spacer()
@@ -453,6 +465,46 @@ struct PlaybackSettings: View {
        }
        .disabled(captions.isEmpty)
    }
+
+    @ViewBuilder private var audioTrackButton: some View {
+        #if os(macOS)
+            audioTrackPicker
+                .labelsHidden()
+                .frame(maxWidth: 300)
+        #elseif os(iOS)
+            Menu {
+                audioTrackPicker
+            } label: {
+                Text(player.availableAudioTracks[player.selectedAudioTrackIndex].displayLanguage)
+                    .frame(maxWidth: 240, alignment: .trailing)
+            }
+            .transaction { t in t.animation = .none }
+            .buttonStyle(.plain)
+            .foregroundColor(.accentColor)
+            .frame(maxWidth: 240, alignment: .trailing)
+            .frame(height: 40)
+        #else
+            ControlsOverlayButton(focusedField: $focusedField, field: .audioTrack) {
+                Text(player.availableAudioTracks[player.selectedAudioTrackIndex].displayLanguage)
+                    .frame(maxWidth: 320)
+            }
+            .contextMenu {
+                ForEach(Array(player.availableAudioTracks.enumerated()), id: \.offset) { index, track in
+                    Button(track.description) { player.selectedAudioTrackIndex = index }
+                }
+                Button("Cancel", role: .cancel) {}
+            }
+        #endif
+    }
+
+    private var audioTrackPicker: some View {
+        Picker("", selection: $player.selectedAudioTrackIndex) {
+            ForEach(Array(player.availableAudioTracks.enumerated()), id: \.offset) { index, track in
+                Text(track.description).tag(index)
+            }
+        }
+        .transaction { t in t.animation = .none }
+    }
 }

 struct PlaybackSettings_Previews: PreviewProvider {