Change the TTS voice

Author: mickael-menu

Sources/Navigator/TTS/AVTTSEngine.swift

@@ -73,10 +73,19 @@ public class AVTTSEngine: NSObject, TTSEngine, AVSpeechSynthesizerDelegate, Logg
         avUtterance.pitchMultiplier = Float(avPitchRange.valueForPercentage(config.pitch))
         avUtterance.preUtteranceDelay = utterance.delay
         avUtterance.postUtteranceDelay = config.delay
-        avUtterance.voice = AVSpeechSynthesisVoice(language: utterance.language ?? config.defaultLanguage)
+        avUtterance.voice = voice(for: utterance)
         return avUtterance
     }
 
+    private func voice(for utterance: TTSUtterance) -> AVSpeechSynthesisVoice? {
+        let language = utterance.language ?? config.defaultLanguage
+        if let voice = config.voice, voice.language == language {
+            return AVSpeechSynthesisVoice(identifier: voice.identifier)
+        } else {
+            return AVSpeechSynthesisVoice(language: language)
+        }
+    }
+
     public func speechSynthesizer(_ synthesizer: AVSpeechSynthesizer, didFinish utterance: AVSpeechUtterance) {
         guard let utterance = (utterance as? AVUtterance)?.utterance else {
             return

Sources/Navigator/TTS/TTSController.swift

@@ -74,8 +74,8 @@ public class TTSController: Loggable, TTSEngineDelegate {
             )
         }
 
-        if let language = publication.metadata.languages.first {
-            engine.config.defaultLanguage = Language(code: .bcp47(language))
+        if let language = publication.metadata.language {
+            engine.config.defaultLanguage = language
         }
         engine.delegate = self
     }
@@ -260,7 +260,7 @@ public class TTSController: Loggable, TTSEngineDelegate {
         return TTSUtterance(
             text: text,
             locator: locator,
-            language: language,
+            language: language.takeIf { $0 != publication.metadata.language },
             delay: delay
         )
     }

Sources/Navigator/TTS/TTSEngine.swift

@@ -1,5 +1,5 @@
 //
-//  Copyright 2021 Readium Foundation. All rights reserved.
+//  Copyright 2022 Readium Foundation. All rights reserved.
 //  Use of this source code is governed by the BSD-style license
 //  available in the top-level LICENSE file of the project.
 //
@@ -23,7 +23,9 @@ public protocol TTSEngineDelegate: AnyObject {
 }
 
 public struct TTSConfiguration {
-    public var defaultLanguage: Language
+    public var defaultLanguage: Language {
+        didSet { voice = nil }
+    }
     public var rate: Double
     public var pitch: Double
     public var voice: TTSVoice?
@@ -44,12 +46,12 @@ public struct TTSConfiguration {
     }
 }
 
-public struct TTSVoice {
-    public enum Gender {
+public struct TTSVoice: Hashable {
+    public enum Gender: Hashable {
         case female, male, unspecified
     }
 
-    public enum Quality {
+    public enum Quality: Hashable {
         case low, medium, high
     }
 

Sources/Shared/Publication/Metadata.swift

@@ -30,8 +30,9 @@ public struct Metadata: Hashable, Loggable, WarningLogger {
 
     public let modified: Date?
     public let published: Date?
-    // FIXME: Use `Language` instead of raw String
     public let languages: [String]  // BCP 47 tag
+    // Main language of the publication.
+    public let language: Language?
     public let sortAs: String?
     public let subjects: [Subject]
     public let authors: [Contributor]
@@ -104,6 +105,7 @@ public struct Metadata: Hashable, Loggable, WarningLogger {
         self.modified = modified
         self.published = published
         self.languages = languages
+        self.language = languages.first.map { Language(code: .bcp47($0)) }
         self.sortAs = sortAs
         self.subjects = subjects
         self.authors = authors
@@ -152,6 +154,7 @@ public struct Metadata: Hashable, Loggable, WarningLogger {
         self.modified = parseDate(json.pop("modified"))
         self.published = parseDate(json.pop("published"))
         self.languages = parseArray(json.pop("language"), allowingSingle: true)
+        self.language = languages.first.map { Language(code: .bcp47($0)) }
         self.sortAs = json.pop("sortAs") as? String
         self.subjects = [Subject](json: json.pop("subject"), warnings: warnings)
         self.authors = [Contributor](json: json.pop("author"), warnings: warnings,  normalizeHREF: normalizeHREF)

Sources/Shared/Publication/Services/Search/StringSearchService.swift

@@ -25,7 +25,7 @@ public class _StringSearchService: _SearchService {
         return { context in
             _StringSearchService(
                 publication: context.publication,
-                language: context.manifest.metadata.languages.first.map { Language(code: .bcp47($0)) },
+                language: context.manifest.metadata.language,
                 snippetLength: snippetLength,
                 searchAlgorithm: searchAlgorithm,
                 extractorFactory: extractorFactory

Sources/Shared/Toolkit/Language.swift

@@ -39,4 +39,10 @@ public struct Language: Hashable {
     public init(locale: Locale) {
         self.init(code: .bcp47(locale.identifier))
     }
+}
+
+extension Language: CustomStringConvertible {
+    public var description: String {
+        code.bcp47
+    }
 }

TestApp/Sources/Reader/Common/TTS/TTSView.swift

@@ -1,5 +1,5 @@
 //
-//  Copyright 2021 Readium Foundation. All rights reserved.
+//  Copyright 2022 Readium Foundation. All rights reserved.
 //  Use of this source code is governed by the BSD-style license
 //  available in the top-level LICENSE file of the project.
 //
@@ -89,6 +89,13 @@ struct TTSSettings: View {
                     choices: viewModel.availableLanguages,
                     choiceLabel: { $0.localizedName }
                 )
+
+                ConfigPicker(
+                    caption: "Voice",
+                    for: \.voice,
+                    choices: viewModel.availableVoices(for: viewModel.config.defaultLanguage),
+                    choiceLabel: { $0?.name ?? "Default" }
+                )
             }
         }
         .listStyle(.insetGrouped)

TestApp/Sources/Reader/Common/TTS/TTSViewModel.swift

@@ -58,6 +58,10 @@ final class TTSViewModel: ObservableObject, Loggable {
 
     var defaultConfig: TTSConfiguration { tts.defaultConfig }
 
+    func availableVoices(for language: Language) -> [TTSVoice?] {
+        [nil] + tts.availableVoices.filter { $0.language == language }
+    }
+
     lazy var availableLanguages: [Language] =
         tts.availableVoices
             .map { $0.language }