add chinese subtitles

2026-10-03 12:14:17 +08:00
parent 138d446d73
commit 6bd64ee4a9
22 changed files with 922 additions and 77 deletions
--- a/ios/OralTrainer/Config.swift
+++ b/ios/OralTrainer/Config.swift
@@ -15,6 +15,7 @@ enum AppConfig {
    static let connectTimeout: TimeInterval = 15
    static let readTimeout: TimeInterval = 30
    static let assessmentTimeout: TimeInterval = 180
+    static let translationTimeout: TimeInterval = 120

    static let sha256Pattern = #"^[0-9a-fA-F]{64}$"#

--- a/ios/OralTrainer/Models/TrainingModels.swift
+++ b/ios/OralTrainer/Models/TrainingModels.swift
@@ -7,6 +7,7 @@ struct SentenceBoundary: Identifiable, Equatable, Decodable {
    let text: String?
    let language: String?
    let referenceSpeechDurationMs: Int64?
+    let zhText: String?

    var id: Int { index }

@@ -15,6 +16,48 @@ struct SentenceBoundary: Identifiable, Equatable, Decodable {
        case startMs = "start_ms"
        case endMs = "end_ms"
        case referenceSpeechDurationMs = "reference_speech_duration_ms"
+        case zhText = "zh_text"
+    }
+
+    init(from decoder: Decoder) throws {
+        let container = try decoder.container(keyedBy: CodingKeys.self)
+        index = try container.decode(Int.self, forKey: .index)
+        startMs = try container.decode(Int64.self, forKey: .startMs)
+        endMs = try container.decode(Int64.self, forKey: .endMs)
+        text = try container.decodeIfPresent(String.self, forKey: .text)
+        language = try container.decodeIfPresent(String.self, forKey: .language)
+        referenceSpeechDurationMs = try container.decodeIfPresent(Int64.self, forKey: .referenceSpeechDurationMs)
+        zhText = try container.decodeIfPresent(String.self, forKey: .zhText)
+    }
+
+    init(
+        index: Int,
+        startMs: Int64,
+        endMs: Int64,
+        text: String?,
+        language: String?,
+        referenceSpeechDurationMs: Int64?,
+        zhText: String? = nil
+    ) {
+        self.index = index
+        self.startMs = startMs
+        self.endMs = endMs
+        self.text = text
+        self.language = language
+        self.referenceSpeechDurationMs = referenceSpeechDurationMs
+        self.zhText = zhText
+    }
+
+    func withZhText(_ zhText: String?) -> SentenceBoundary {
+        SentenceBoundary(
+            index: index,
+            startMs: startMs,
+            endMs: endMs,
+            text: text,
+            language: language,
+            referenceSpeechDurationMs: referenceSpeechDurationMs,
+            zhText: zhText
+        )
    }
 }

--- a/ios/OralTrainer/Services/SentenceBoundaryAPI.swift
+++ b/ios/OralTrainer/Services/SentenceBoundaryAPI.swift
@@ -46,6 +46,26 @@ struct SentenceExplanationResponse: Decodable {
    }
 }

+struct SubtitleTranslationItem: Decodable {
+    let index: Int
+    let zhText: String
+
+    private enum CodingKeys: String, CodingKey {
+        case index
+        case zhText = "zh_text"
+    }
+}
+
+struct SubtitleTranslationResponse: Decodable {
+    let videoHash: String
+    let translations: [SubtitleTranslationItem]
+
+    private enum CodingKeys: String, CodingKey {
+        case translations
+        case videoHash = "video_hash"
+    }
+}
+
 extension SentenceBoundaryAPI {
    static func explain(
        _ sentence: SentenceBoundary,
@@ -79,4 +99,32 @@ extension SentenceBoundaryAPI {
            .decode(SentenceExplanationResponse.self, from: result.body)
            .explanation
    }
+
+    static func translateSubtitles(videoHash: String) async throws -> [Int: String] {
+        let hash = videoHash.lowercased()
+        guard hash.range(of: AppConfig.sha256Pattern, options: .regularExpression) != nil else {
+            throw SentenceBoundaryAPIError.invalidHash
+        }
+        var headers: [String: String] = ["Accept": "application/json"]
+        if !AppConfig.authToken.isEmpty {
+            headers["Authorization"] = "Bearer \(AppConfig.authToken)"
+        }
+        let result = try await HTTPClient.execute(
+            baseURL: AppConfig.serverBaseURL,
+            path: "api/v1/videos/\(hash)/subtitle-translation",
+            method: "POST",
+            timeout: AppConfig.translationTimeout,
+            headers: headers
+        )
+        guard (200...299).contains(result.statusCode) else {
+            let detail = String(data: result.body, encoding: .utf8) ?? ""
+            throw HTTPClientError.badStatus(result.statusCode, detail)
+        }
+        let payload = try JSONDecoder().decode(SubtitleTranslationResponse.self, from: result.body)
+        var mapped: [Int: String] = [:]
+        for item in payload.translations where !item.zhText.isEmpty {
+            mapped[item.index] = item.zhText
+        }
+        return mapped
+    }
 }
--- a/ios/OralTrainer/Views/FullscreenPlayerView.swift
+++ b/ios/OralTrainer/Views/FullscreenPlayerView.swift
@@ -4,6 +4,8 @@ import SwiftUI
 struct FullscreenPlayerView: View {
    @ObservedObject var controller: PlayerController
    var showsLessonTitle: Bool = true
+    var showChineseSubtitles: Bool = false
+    var chineseSubtitlePlaceholder: String = "正在翻译中文字幕…"
    let onExit: () -> Void

    @State private var subtitleVisible = true
@@ -66,10 +68,18 @@ struct FullscreenPlayerView: View {
                            .font(.system(size: 13))
                            .foregroundColor(Theme.textOverlay)
                        HStack(alignment: .center, spacing: 8) {
-                            Text(sentenceTextText)
-                            .font(.system(size: 15))
-                            .foregroundColor(.white)
-                            .lineLimit(2)
+                            VStack(alignment: .leading, spacing: 4) {
+                                Text(sentenceTextText)
+                                    .font(.system(size: 15))
+                                    .foregroundColor(.white)
+                                    .lineLimit(2)
+                                if let chineseSubtitleText = chineseSubtitleText {
+                                    Text(chineseSubtitleText)
+                                        .font(.system(size: 14))
+                                        .foregroundColor(Theme.textOverlay)
+                                        .lineLimit(2)
+                                }
+                            }
                            Spacer()
                            Button {
                                guard let sentence = controller.currentSentence,
@@ -203,6 +213,20 @@ struct FullscreenPlayerView: View {
        }
    }

+    private var chineseSubtitleText: String? {
+        guard showChineseSubtitles else {
+            return nil
+        }
+        if let zh = controller.currentSentence?.zhText?.trimmingCharacters(in: .whitespacesAndNewlines),
+           !zh.isEmpty {
+            return zh
+        }
+        if (controller.currentSentence?.text ?? "").trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
+            return nil
+        }
+        return chineseSubtitlePlaceholder
+    }
+
    private var statusText: String {
        if let message = controller.cacheState.statusMessage {
            return message
--- a/ios/OralTrainer/Views/PlayerSectionView.swift
+++ b/ios/OralTrainer/Views/PlayerSectionView.swift
@@ -7,6 +7,8 @@ struct PlayerSectionView: View {
    @Binding var continuousPlayback: Bool
    var showsSubtitle: Binding<Bool>? = .constant(true)
    var showsLessonTitle: Bool = true
+    var showChineseSubtitles: Binding<Bool>? = nil
+    var chineseSubtitlePlaceholder: String = "正在翻译中文字幕…"
    var onFullscreen: () -> Void

    @State private var progressDragging = false
@@ -93,14 +95,34 @@ struct PlayerSectionView: View {
                }
                .padding(.vertical, 2)
            }
+            if let showChineseSubtitles = showChineseSubtitles {
+                HStack(spacing: 0) {
+                    Text("中文字幕")
+                        .font(.system(size: 13, weight: .bold))
+                        .foregroundColor(Theme.textSubtle)
+                    Spacer()
+                    Toggle("", isOn: showChineseSubtitles)
+                        .labelsHidden()
+                        .tint(Theme.accent)
+                }
+                .padding(.vertical, 2)
+            }
            Text(sentenceMetaText)
                .font(.system(size: 13))
                .foregroundColor(Theme.textSubtle)
            HStack(alignment: .center, spacing: 8) {
-                Text(sentenceTextText)
-                .font(.system(size: 15))
-                    .foregroundColor(.white)
-                    .lineLimit(4)
+                VStack(alignment: .leading, spacing: 4) {
+                    Text(sentenceTextText)
+                        .font(.system(size: 15))
+                        .foregroundColor(.white)
+                        .lineLimit(4)
+                    if let chineseSubtitleText = chineseSubtitleText {
+                        Text(chineseSubtitleText)
+                            .font(.system(size: 14))
+                            .foregroundColor(Theme.textMuted)
+                            .lineLimit(4)
+                    }
+                }
                Spacer()
                Button {
                    guard let sentence = controller.currentSentence, !(sentence.text ?? "").isEmpty else {
@@ -205,6 +227,20 @@ struct PlayerSectionView: View {
        }
    }

+    private var chineseSubtitleText: String? {
+        guard showChineseSubtitles?.wrappedValue == true else {
+            return nil
+        }
+        if let zh = controller.currentSentence?.zhText?.trimmingCharacters(in: .whitespacesAndNewlines),
+           !zh.isEmpty {
+            return zh
+        }
+        if (controller.currentSentence?.text ?? "").trimmingCharacters(in: .whitespacesAndNewlines).isEmpty {
+            return nil
+        }
+        return chineseSubtitlePlaceholder
+    }
+
    private var statusText: String {
        if let message = controller.cacheState.statusMessage {
            return message
--- a/ios/OralTrainer/Views/RootView.swift
+++ b/ios/OralTrainer/Views/RootView.swift
@@ -9,6 +9,10 @@ struct RootView: View {
    @StateObject private var historyManager = TrainingHistoryManager()
    @State private var activeModule: Module = .course
    @State private var continuousPlayback = false
+    @State private var showChineseSubtitles = false
+    @State private var chineseTranslationLoading = false
+    @State private var chineseTranslationError = ""
+    @State private var chineseTranslationTask: Task<Void, Never>?
    @State private var showImporter = false
    @State private var showFullscreenPlayer = false
    @State private var courses: [Course] = []
@@ -53,6 +57,8 @@ struct RootView: View {
                        controller: controller,
                        catalogVM: catalogVM,
                        continuousPlayback: $continuousPlayback,
+                        showChineseSubtitles: $showChineseSubtitles,
+                        chineseSubtitlePlaceholder: chineseSubtitlePlaceholder,
                        onFullscreen: enterFullscreen
                    )
                case .test:
@@ -88,9 +94,24 @@ struct RootView: View {
                Task { await loadCourses() }
            }
        }
+        .onChange(of: showChineseSubtitles) { enabled in
+            if enabled {
+                ensureChineseTranslations()
+            } else {
+                chineseTranslationTask?.cancel()
+                chineseTranslationLoading = false
+            }
+        }
+        .onChange(of: controller.currentItem?.id) { _ in
+            ensureChineseTranslations()
+        }
+        .onChange(of: controller.currentItem?.sentences.count) { _ in
+            ensureChineseTranslations()
+        }
        .onChange(of: activeModule) { module in
            if module == .train {
                controller.setContinuousPlayback(continuousPlayback)
+                ensureChineseTranslations()
            } else if module == .test {
                // 与安卓测试页一致：按当前句窗口播放，避免滑进下一句。
                controller.setContinuousPlayback(false)
@@ -115,7 +136,9 @@ struct RootView: View {
        .fullScreenCover(isPresented: $showFullscreenPlayer) {
            FullscreenPlayerView(
                controller: controller,
-                showsLessonTitle: activeModule != .train
+                showsLessonTitle: activeModule != .train,
+                showChineseSubtitles: showChineseSubtitles,
+                chineseSubtitlePlaceholder: chineseSubtitlePlaceholder
            ) {
                exitFullscreen()
            }
@@ -593,4 +616,59 @@ struct RootView: View {
        AppDelegate.lockOrientation(restored)
        showFullscreenPlayer = false
    }
+
+    private var chineseSubtitlePlaceholder: String {
+        if chineseTranslationLoading {
+            return "正在翻译中文字幕…"
+        }
+        if !chineseTranslationError.isEmpty {
+            return chineseTranslationError
+        }
+        return "暂无中文字幕"
+    }
+
+    private func ensureChineseTranslations() {
+        guard showChineseSubtitles, activeModule == .train else {
+            return
+        }
+        guard let item = controller.currentItem else {
+            return
+        }
+        let needsTranslation = item.sentences.contains {
+            !($0.text ?? "").trimmingCharacters(in: .whitespacesAndNewlines).isEmpty &&
+                ($0.zhText ?? "").trimmingCharacters(in: .whitespacesAndNewlines).isEmpty
+        }
+        guard needsTranslation else {
+            chineseTranslationLoading = false
+            chineseTranslationError = ""
+            return
+        }
+        chineseTranslationTask?.cancel()
+        chineseTranslationLoading = true
+        chineseTranslationError = ""
+        let mediaId = item.id
+        chineseTranslationTask = Task {
+            do {
+                let map = try await SentenceBoundaryAPI.translateSubtitles(videoHash: mediaId)
+                guard !Task.isCancelled else { return }
+                await MainActor.run {
+                    guard controller.currentItem?.id == mediaId else { return }
+                    let updated = (controller.currentItem?.sentences ?? []).map { sentence in
+                        sentence.withZhText(map[sentence.index] ?? sentence.zhText)
+                    }
+                    controller.updateBoundaries(updated)
+                    chineseTranslationLoading = false
+                    chineseTranslationError = ""
+                    controller.showTransientMessage("已加载中文字幕")
+                }
+            } catch {
+                guard !Task.isCancelled else { return }
+                await MainActor.run {
+                    chineseTranslationLoading = false
+                    chineseTranslationError = "中文字幕翻译失败"
+                    controller.showTransientMessage("中文字幕翻译失败：\(error.localizedDescription)")
+                }
+            }
+        }
+    }
 }
--- a/ios/OralTrainer/Views/TrainingView.swift
+++ b/ios/OralTrainer/Views/TrainingView.swift
@@ -4,6 +4,8 @@ struct TrainingView: View {
    @ObservedObject var controller: PlayerController
    @ObservedObject var catalogVM: CatalogViewModel
    @Binding var continuousPlayback: Bool
+    @Binding var showChineseSubtitles: Bool
+    var chineseSubtitlePlaceholder: String = "正在翻译中文字幕…"
    let onFullscreen: () -> Void

    var body: some View {
@@ -14,6 +16,8 @@ struct TrainingView: View {
                    showsContinuousToggle: true,
                    continuousPlayback: $continuousPlayback,
                    showsLessonTitle: false,
+                    showChineseSubtitles: $showChineseSubtitles,
+                    chineseSubtitlePlaceholder: chineseSubtitlePlaceholder,
                    onFullscreen: onFullscreen
                )
                CatalogView(vm: catalogVM)
@@ -28,5 +32,8 @@ struct TrainingView: View {
                    : "已关闭连续播放，将循环播放当前句子"
            )
        }
+        .onChange(of: showChineseSubtitles) { enabled in
+            controller.showTransientMessage(enabled ? "已开启中文字幕" : "已关闭中文字幕")
+        }
    }
 }