fixed some bugs on ios

ios/OralTrainer.xcodeproj/project.pbxproj

@@ -29,6 +29,7 @@
 		AA0000000000000000000054 /* FullscreenPlayerView.swift in Sources */ = {isa = PBXBuildFile; fileRef = AA0000000000000000000034 /* FullscreenPlayerView.swift */; };
 		AA0000000000000000000060 /* InfoPlist.strings in Resources */ = {isa = PBXBuildFile; fileRef = AA0000000000000000000070 /* InfoPlist.strings */; };
 		AA000000000000000000007D /* DubbingCompositor.swift in Sources */ = {isa = PBXBuildFile; fileRef = AA000000000000000000007E /* DubbingCompositor.swift */; };
+		AA0000000000000000000080 /* AudioRecorderSession.swift in Sources */ = {isa = PBXBuildFile; fileRef = AA0000000000000000000081 /* AudioRecorderSession.swift */; };
 /* End PBXBuildFile section */
 
 /* Begin PBXFileReference section */
@@ -56,6 +57,7 @@
 		AA0000000000000000000034 /* FullscreenPlayerView.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = FullscreenPlayerView.swift; sourceTree = "<group>"; };
 		AA0000000000000000000071 /* zh-Hans */ = {isa = PBXFileReference; lastKnownFileType = text.plist.strings; name = "zh-Hans"; path = "zh-Hans.lproj/InfoPlist.strings"; sourceTree = "<group>"; };
 		AA000000000000000000007E /* DubbingCompositor.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = DubbingCompositor.swift; sourceTree = "<group>"; };
+		AA0000000000000000000081 /* AudioRecorderSession.swift */ = {isa = PBXFileReference; lastKnownFileType = sourcecode.swift; path = AudioRecorderSession.swift; sourceTree = "<group>"; };
 /* End PBXFileReference section */
 
 /* Begin PBXFrameworksBuildPhase section */
@@ -118,6 +120,7 @@
 				AA000000000000000000002A /* SentenceBoundaryAPI.swift */,
 				AA000000000000000000002B /* ImitationAssessor.swift */,
 				AA000000000000000000007E /* DubbingCompositor.swift */,
+				AA0000000000000000000081 /* AudioRecorderSession.swift */,
 			);
 			path = Services;
 			sourceTree = "<group>";
@@ -237,6 +240,7 @@
 				AA000000000000000000004A /* SentenceBoundaryAPI.swift in Sources */,
 				AA000000000000000000004B /* ImitationAssessor.swift in Sources */,
 				AA000000000000000000007D /* DubbingCompositor.swift in Sources */,
+				AA0000000000000000000080 /* AudioRecorderSession.swift in Sources */,
 				AA000000000000000000004C /* PlayerController.swift in Sources */,
 				AA000000000000000000004D /* PlayerSurfaceView.swift in Sources */,
 				AA000000000000000000004E /* Theme.swift in Sources */,

ios/OralTrainer/Services/AudioRecorderSession.swift

@@ -0,0 +1,159 @@
+import AVFoundation
+import Foundation
+
+enum AudioRecorderSessionError: LocalizedError {
+    case inputUnavailable
+    case startFailed
+    case conversionFailed
+
+    var errorDescription: String? {
+        switch self {
+        case .inputUnavailable:
+            return "未检测到麦克风输入"
+        case .startFailed:
+            return "无法开始录制：请确认麦克风未被其他应用占用后重试"
+        case .conversionFailed:
+            return "录音文件转码失败"
+        }
+    }
+}
+
+/// 基于 AVAudioEngine 的录音会话。
+///
+/// AVAudioRecorder 在 iOS 26 真机上存在 record() 恒返回 false 的已知问题
+/// （Apple 开发者论坛多个报告），因此这里改用 AVAudioEngine 采集 PCM，
+/// 停止录音时再转码为 16kHz 单声道 AAC（m4a），与服务端评测接口保持一致。
+final class AudioRecorderSession {
+    private let engine = AVAudioEngine()
+    private let outputURL: URL
+    private let pcmURL: URL
+    private var inputFile: AVAudioFile?
+    private var tapInstalled = false
+
+    private(set) var isRunning = false
+    private(set) var recordDuration: TimeInterval = 0
+
+    /// 本次录音的原始 PCM 文件（caf），停止录音后仍保留，
+    /// 供配音合成直接合并，避免在 iOS 26 上读回 AAC（AVAudioFile 存在 'fmt?' 问题）。
+    var rawRecordingURL: URL { pcmURL }
+
+    init(outputURL: URL) {
+        self.outputURL = outputURL
+        let stamp = Int(Date().timeIntervalSince1970 * 1000)
+        self.pcmURL = FileManager.default.temporaryDirectory
+            .appendingPathComponent("raw-\(stamp).caf")
+    }
+
+    /// 激活音频会话并开始采集。失败时抛出错误，可在重新激活会话后重试。
+    func start() throws {
+        let session = AVAudioSession.sharedInstance()
+        try session.setCategory(.playAndRecord, mode: .default, options: [.defaultToSpeaker])
+        try session.setActive(true)
+        guard session.isInputAvailable else {
+            throw AudioRecorderSessionError.inputUnavailable
+        }
+
+        let input = engine.inputNode
+        if !tapInstalled {
+            let inputFormat = input.outputFormat(forBus: 0)
+            guard inputFormat.sampleRate > 0, inputFormat.channelCount > 0 else {
+                throw AudioRecorderSessionError.inputUnavailable
+            }
+            let file = try AVAudioFile(forWriting: pcmURL, settings: inputFormat.settings)
+            inputFile = file
+            input.installTap(onBus: 0, bufferSize: 4096, format: inputFormat) { [weak self] buffer, _ in
+                guard let self = self else { return }
+                do {
+                    try self.inputFile?.write(from: buffer)
+                } catch {
+                    // 单个缓冲写入失败时忽略，继续录音
+                }
+            }
+            tapInstalled = true
+        }
+        engine.prepare()
+        do {
+            try engine.start()
+        } catch {
+            throw AudioRecorderSessionError.startFailed
+        }
+        isRunning = true
+    }
+
+    /// 停止录音，将 PCM 转码为 m4a 写入 outputURL。
+    func stop() throws {
+        guard isRunning else { return }
+        engine.inputNode.removeTap(onBus: 0)
+        engine.stop()
+        isRunning = false
+        tapInstalled = false
+        inputFile = nil
+        try finalizeToAAC()
+    }
+
+    /// 放弃本次录音并清理临时文件。
+    func cancel() {
+        if isRunning {
+            engine.inputNode.removeTap(onBus: 0)
+            engine.stop()
+            isRunning = false
+            tapInstalled = false
+            inputFile = nil
+        }
+        try? FileManager.default.removeItem(at: pcmURL)
+        try? FileManager.default.removeItem(at: outputURL)
+    }
+
+    private func finalizeToAAC() throws {
+        let inputFile = try AVAudioFile(forReading: pcmURL)
+        recordDuration = Double(inputFile.length) / inputFile.processingFormat.sampleRate
+        let outputSettings: [String: Any] = [
+            AVFormatIDKey: kAudioFormatMPEG4AAC,
+            AVSampleRateKey: 16_000,
+            AVNumberOfChannelsKey: 1,
+        ]
+        let outputFile = try AVAudioFile(forWriting: outputURL, settings: outputSettings)
+        guard let converter = AVAudioConverter(from: inputFile.processingFormat, to: outputFile.processingFormat) else {
+            throw AudioRecorderSessionError.conversionFailed
+        }
+        let inputBlock: AVAudioConverterInputBlock = { inNumPackets, outStatus in
+            if inputFile.framePosition >= inputFile.length {
+                outStatus.pointee = .endOfStream
+                return nil
+            }
+            let remaining = inputFile.length - inputFile.framePosition
+            let frames = AVAudioFrameCount(min(inNumPackets, AVAudioFrameCount(remaining)))
+            guard frames > 0,
+                  let buffer = AVAudioPCMBuffer(pcmFormat: inputFile.processingFormat, frameCapacity: frames) else {
+                outStatus.pointee = .noDataNow
+                return nil
+            }
+            do {
+                try inputFile.read(into: buffer)
+                outStatus.pointee = .haveData
+                return buffer
+            } catch {
+                outStatus.pointee = .noDataNow
+                return nil
+            }
+        }
+        let capacity = AVAudioFrameCount(min(4096, max(1, Int(inputFile.length))))
+        guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: outputFile.processingFormat, frameCapacity: capacity) else {
+            throw AudioRecorderSessionError.conversionFailed
+        }
+        while true {
+            var conversionError: NSError?
+            let status = converter.convert(to: outputBuffer, error: &conversionError, withInputFrom: inputBlock)
+            switch status {
+            case .haveData:
+                try outputFile.write(from: outputBuffer)
+            case .inputRanDry, .endOfStream:
+                return
+            case .error:
+                throw conversionError ?? AudioRecorderSessionError.conversionFailed
+            default:
+                return
+            }
+        }
+    }
+}

ios/OralTrainer/Services/DubbingCompositor.swift

@@ -2,61 +2,153 @@ import AVFoundation
 import Foundation
 
 enum DubbingCompositorError: LocalizedError {
-    case noAudioTracks
+    case inputReadFailed(String?)
-    case cannotReadDuration
+    case outputWriteFailed(String?)
+    case convertFailed(String?)
 
     var errorDescription: String? {
         switch self {
-        case .noAudioTracks:
+        case .inputReadFailed(let detail):
-            return "录音没有音频轨道"
+            return detail.map { "无法读取录音音频（\($0)）" } ?? "无法读取录音音频"
-        case .cannotReadDuration:
+        case .outputWriteFailed(let detail):
-            return "无法读取录音时长"
+            return detail.map { "无法写入合成音频（\($0)）" } ?? "无法写入合成音频"
+        case .convertFailed(let detail):
+            return detail.map { "无法转换音频格式（\($0)）" } ?? "无法转换音频格式"
         }
     }
 }
 
 enum DubbingCompositor {
+    /// 合并多段录音为一段 16kHz 单声道 AAC（m4a）。
+    ///
+    /// 不使用 AVMutableComposition + AVAssetExportSession：iOS 26 / macOS 26 上
+    /// AVAssetExportSession 对纯音频资源导出会失败（AVError -11800，底层 'fmt?'），
+    /// AVAudioFile 读回 AAC 也会失败（CheckClientFormatSet / 'fmt?'）。
+    /// 因此传入录音时保留的原始 PCM（caf），解码拼接后直接编码 AAC，
+    /// 与录音转码走同一条已验证可用的链路。
     static func merge(urls: [URL], outputURL: URL) async throws {
-        try await mergeSync(urls: urls, outputURL: outputURL)
+        do {
+            try mergeSync(urls: urls, outputURL: outputURL)
+        } catch {
+            try? FileManager.default.removeItem(at: outputURL)
+            throw error
+        }
     }
 
-    private static func mergeSync(urls: [URL], outputURL: URL) async throws {
+    private static func mergeSync(urls: [URL], outputURL: URL) throws {
         try? FileManager.default.removeItem(at: outputURL)
-        let composition = AVMutableComposition()
+        let settings: [String: Any] = [
-        var currentTime = CMTime.zero
+            AVFormatIDKey: kAudioFormatMPEG4AAC,
+            AVSampleRateKey: 16_000,
+            AVNumberOfChannelsKey: 1,
+        ]
+        let outputFile: AVAudioFile
+        do {
+            outputFile = try AVAudioFile(forWriting: outputURL, settings: settings)
+        } catch {
+            throw DubbingCompositorError.outputWriteFailed(error.localizedDescription)
+        }
+        let targetFormat = outputFile.processingFormat
         for url in urls {
-            let source = AVURLAsset(url: url)
+            let inputFile: AVAudioFile
-            let tracks = try await source.loadTracks(withMediaType: .audio)
+            do {
-            guard let track = tracks.first else {
+                inputFile = try AVAudioFile(forReading: url)
-                throw DubbingCompositorError.noAudioTracks
+            } catch {
+                throw DubbingCompositorError.inputReadFailed(error.localizedDescription)
             }
-            let duration = try await source.load(.duration)
+            guard inputFile.length > 0 else {
-            guard duration.seconds.isFinite, duration.seconds > 0 else {
+                throw DubbingCompositorError.inputReadFailed("录音为空")
-                throw DubbingCompositorError.cannotReadDuration
             }
-            let timeRange = CMTimeRange(start: .zero, duration: duration)
+            let inputFormat = inputFile.processingFormat
-            let compositionTrack = composition.addMutableTrack(
+            if inputFormat.sampleRate == targetFormat.sampleRate,
-                withMediaType: .audio,
+               inputFormat.channelCount == targetFormat.channelCount {
-                preferredTrackID: kCMPersistentTrackID_Invalid
+                try appendStream(from: inputFile, to: outputFile)
-            )
+            } else {
-            guard let compositionTrack else {
+                try appendConverted(from: inputFile, to: outputFile, targetFormat: targetFormat)
-                throw DubbingCompositorError.noAudioTracks
             }
-            try compositionTrack.insertTimeRange(timeRange, of: track, at: currentTime)
-            currentTime = currentTime + duration
         }
+    }
 
-        guard let exporter = AVAssetExportSession(
+    private static func appendStream(from inputFile: AVAudioFile, to outputFile: AVAudioFile) throws {
-            asset: composition,
+        var remaining = inputFile.length
-            presetName: AVAssetExportPresetAppleM4A
+        while remaining > 0 {
-        ) else {
+            let frames = AVAudioFrameCount(min(remaining, 8192))
-            throw DubbingCompositorError.noAudioTracks
+            guard let buffer = AVAudioPCMBuffer(pcmFormat: inputFile.processingFormat, frameCapacity: frames) else {
+                throw DubbingCompositorError.convertFailed("无法分配解码缓冲")
+            }
+            do {
+                try inputFile.read(into: buffer)
+            } catch {
+                throw DubbingCompositorError.inputReadFailed(error.localizedDescription)
+            }
+            guard buffer.frameLength > 0 else {
+                throw DubbingCompositorError.inputReadFailed("录音文件读取出错（读取到空帧）")
+            }
+            do {
+                try outputFile.write(from: buffer)
+            } catch {
+                throw DubbingCompositorError.outputWriteFailed(error.localizedDescription)
+            }
+            remaining -= AVAudioFramePosition(buffer.frameLength)
         }
-        exporter.outputURL = outputURL
+    }
-        exporter.outputFileType = .m4a
+
-        await exporter.export()
+    private static func appendConverted(
-        guard exporter.status == .completed else {
+        from inputFile: AVAudioFile,
-            throw exporter.error ?? DubbingCompositorError.noAudioTracks
+        to outputFile: AVAudioFile,
+        targetFormat: AVAudioFormat
+    ) throws {
+        guard let converter = AVAudioConverter(from: inputFile.processingFormat, to: targetFormat) else {
+            throw DubbingCompositorError.convertFailed("无法创建格式转换器")
+        }
+        let inputBlock: AVAudioConverterInputBlock = { inNumPackets, outStatus in
+            if inputFile.framePosition >= inputFile.length {
+                outStatus.pointee = .endOfStream
+                return nil
+            }
+            let remaining = inputFile.length - inputFile.framePosition
+            let frames = AVAudioFrameCount(min(inNumPackets, AVAudioFrameCount(remaining)))
+            guard frames > 0,
+                  let buffer = AVAudioPCMBuffer(pcmFormat: inputFile.processingFormat, frameCapacity: frames) else {
+                outStatus.pointee = .noDataNow
+                return nil
+            }
+            do {
+                try inputFile.read(into: buffer)
+                guard buffer.frameLength > 0 else {
+                    outStatus.pointee = .endOfStream
+                    return nil
+                }
+                outStatus.pointee = .haveData
+                return buffer
+            } catch {
+                outStatus.pointee = .noDataNow
+                return nil
+            }
+        }
+        guard let outputBuffer = AVAudioPCMBuffer(pcmFormat: targetFormat, frameCapacity: 8192) else {
+            throw DubbingCompositorError.convertFailed("无法分配转码缓冲")
+        }
+        while true {
+            var error: NSError?
+            let status = converter.convert(to: outputBuffer, error: &error, withInputFrom: inputBlock)
+            switch status {
+            case .haveData:
+                do {
+                    try outputFile.write(from: outputBuffer)
+                } catch {
+                    throw DubbingCompositorError.outputWriteFailed(error.localizedDescription)
+                }
+            case .inputRanDry, .endOfStream:
+                return
+            case .error:
+                if let error {
+                    throw DubbingCompositorError.convertFailed(error.localizedDescription)
+                }
+                throw DubbingCompositorError.convertFailed(nil)
+            default:
+                return
+            }
         }
     }
 }

ios/OralTrainer/Views/TestView.swift

@@ -5,7 +5,7 @@ struct TestView: View {
     @ObservedObject var controller: PlayerController
     let onFullscreen: () -> Void
 
-    @State private var recorder: AVAudioRecorder?
+    @State private var recordingSession: AudioRecorderSession?
     @State private var recordingURL: URL?
     @State private var isRecording = false
     @State private var isAssessing = false
@@ -13,6 +13,7 @@ struct TestView: View {
     @State private var scoreSummary = "尚未评测"
     @State private var scoreDetail = ""
     @State private var dubSegments: [Int: URL] = [:]
+    @State private var dubRawSegments: [Int: URL] = [:]
     @State private var latestScores: [Int: ImitationAssessmentResult] = [:]
     @State private var mergedDubURL: URL?
     @State private var dubPlayer: AVPlayer?
@@ -82,7 +83,7 @@ struct TestView: View {
                     dubButton("播放配音", enabled: !dubSegments.isEmpty && !dubbingPlayback && !isGeneratingDub) {
                         toggleDubPlayback()
                     }
-                    dubButton("生成配音", enabled: !dubSegments.isEmpty && !dubbingPlayback && !isGeneratingDub) {
+                    dubButton("生成配音", enabled: !dubRawSegments.isEmpty && !dubbingPlayback && !isGeneratingDub) {
                         generateDubShareAudio()
                     }
                     dubButton("分享配音", enabled: mergedDubURL != nil && !dubbingPlayback && !isSharingDub) {
@@ -195,58 +196,99 @@ struct TestView: View {
     private func beginRecording() {
         let fileURL = FileManager.default.temporaryDirectory
             .appendingPathComponent("attempt-\(Int(Date().timeIntervalSince1970 * 1000)).m4a")
-        let settings: [String: Any] = [
+        let newSession = AudioRecorderSession(outputURL: fileURL)
-            AVFormatIDKey: Int(kAudioFormatMPEG4AAC),
-            AVSampleRateKey: 16_000,
-            AVNumberOfChannelsKey: 1,
-            AVEncoderBitRateKey: 96_000,
-            AVEncoderAudioQualityKey: AVAudioQuality.medium.rawValue,
-        ]
         do {
             let session = AVAudioSession.sharedInstance()
             try session.setCategory(.playAndRecord, mode: .default, options: [.defaultToSpeaker])
             try session.setActive(true)
-            let newRecorder = try AVAudioRecorder(url: fileURL, settings: settings)
+            guard session.isInputAvailable else {
-            guard newRecorder.record() else {
+                testStatus = microphoneInputUnavailableMessage()
-                testStatus = "录音启动失败：无法开始录制"
                 return
             }
-            recorder = newRecorder
+            do {
+                try newSession.start()
+            } catch {
+                try session.setActive(false, options: .notifyOthersOnDeactivation)
+                try session.setCategory(.playAndRecord, mode: .default, options: [.defaultToSpeaker])
+                try session.setActive(true)
+                try newSession.start()
+            }
+            recordingSession = newSession
             recordingURL = fileURL
             isRecording = true
             testStatus = "正在录音…读完当前句子后点击“停止并评测”"
+        } catch AudioRecorderSessionError.inputUnavailable {
+            newSession.cancel()
+            testStatus = microphoneInputUnavailableMessage()
+        } catch AudioRecorderSessionError.startFailed {
+            newSession.cancel()
+            testStatus = recordingStartFailedMessage()
         } catch {
+            newSession.cancel()
             testStatus = "录音启动失败：\(error.localizedDescription)"
         }
     }
 
+    private func microphoneInputUnavailableMessage() -> String {
+        #if targetEnvironment(simulator)
+        return "录音启动失败：未检测到麦克风输入。请在模拟器菜单 I/O → Audio Input 中选择输入设备，并允许“模拟器”使用 Mac 麦克风后重试"
+        #else
+        return "录音启动失败：未检测到麦克风输入。请检查 设置 → 隐私与安全性 → 麦克风 是否允许本 App 使用麦克风"
+        #endif
+    }
+
+    private func recordingStartFailedMessage() -> String {
+        #if targetEnvironment(simulator)
+        return "录音启动失败：无法开始录制。模拟器请确认 I/O → Audio Input 已选择输入设备，再重试"
+        #else
+        return "录音启动失败：无法开始录制。请确认麦克风未被其他应用占用后重试"
+        #endif
+    }
+
     private func stopRecordingAndAssess() {
-        guard let currentRecorder = recorder else {
+        guard let currentSession = recordingSession else {
+            return
+        }
+        let duration: TimeInterval
+        do {
+            try currentSession.stop()
+            duration = currentSession.recordDuration
+        } catch {
+            testStatus = "录音结束失败：\(error.localizedDescription)"
+            currentSession.cancel()
+            recordingSession = nil
+            isRecording = false
+            restorePlaybackAudioSession()
             return
         }
-        let duration = currentRecorder.currentTime
-        currentRecorder.stop()
         restorePlaybackAudioSession()
-        recorder = nil
+        recordingSession = nil
         isRecording = false
         guard let url = recordingURL else {
             return
         }
         recordingURL = nil
+        let rawURL = currentSession.rawRecordingURL
         if duration < 1.0 {
             testStatus = "录音太短或无法保存"
             try? FileManager.default.removeItem(at: url)
+            try? FileManager.default.removeItem(at: rawURL)
             return
         }
         guard let sentence = controller.currentSentence else {
             testStatus = "没有可归属的句子，本次录音已丢弃"
             try? FileManager.default.removeItem(at: url)
+            try? FileManager.default.removeItem(at: rawURL)
             return
         }
         if let oldURL = dubSegments[sentence.index] {
             try? FileManager.default.removeItem(at: oldURL)
         }
+        if let oldRawURL = dubRawSegments[sentence.index] {
+            try? FileManager.default.removeItem(at: oldRawURL)
+        }
         dubSegments[sentence.index] = url
+        dubRawSegments[sentence.index] = rawURL
         mergedDubURL = nil
         submitAssessment(recordingURL: url)
     }
@@ -292,7 +334,6 @@ struct TestView: View {
                 scoreSummary = "评测失败"
             }
             isAssessing = false
-            try? FileManager.default.removeItem(at: url)
         }
     }
 
@@ -399,8 +440,8 @@ struct TestView: View {
     }
 
     private func generateDubShareAudio() {
-        guard !dubSegments.isEmpty else { return }
+        guard !dubRawSegments.isEmpty else { return }
-        let segments = dubSegments.sorted { $0.key < $1.key }
+        let segments = dubRawSegments.sorted { $0.key < $1.key }
         let outputURL = FileManager.default.temporaryDirectory
             .appendingPathComponent("dubbing-\(Int(Date().timeIntervalSince1970 * 1000)).m4a")
         isGeneratingDub = true

ios/README.md

@@ -59,4 +59,12 @@ ios/
 
 - “云端暂不可用：…”：检查手机能否访问服务端、地址是否填对、网关证书是否有效。
 - 本地视频评测提示“仅云端课程支持朗读评测”：本地文件需先通过管理端上传到服务器。
-- 录音提示失败：在 设置 > 隐私 > 麦克风 中允许本 App 使用麦克风。
+- 录音提示“无法开始录制”：录音已改用 `AVAudioEngine` 采集并转码 AAC 16kHz
+  （`AVAudioRecorder` 在 iOS 26 真机上存在 `record()` 恒返回 `false` 的已知问题）。
+  若仍失败，真机请确认 设置 → 隐私与安全性 → 麦克风 已允许本 App 使用麦克风，
+  且没有其他 App 正在占用麦克风；模拟器请先在 I/O → Audio Input 中选择输入设备
+  （如 Mac 内置麦克风），并允许“模拟器”使用 Mac 麦克风。
+- 配音“合成失败”：iOS 26 / macOS 26 上 `AVAssetExportSession` 导出纯音频会失败
+  （AVError -11800，底层 `'fmt?'`），`AVAudioFile` 读回 AAC 也会报 `'fmt?'`。
+  配音合成已改为合并录音时保留的原始 PCM（caf）后直接编码 AAC 16kHz，与录音转码
+  同链路；如仍失败，提示中会附带底层错误详情，可按详情排查。