fix: audio recording - double buffer + HLS variant switch support
- Replace single dataBuffer with double-buffer (dataBuffers tuple) to eliminate data race between real-time tapProcess thread and async dispatch queue reading the buffer - Use kCMPersistentTrackID_Invalid for audioMix trackID so the tap survives HLS variant switches (track ID changes on quality change)
This commit is contained in:
parent
1733ecc7bc
commit
fb1bd32bbf
@ -5,151 +5,185 @@ import Foundation
|
||||
#if os(macOS)
|
||||
import AppKit
|
||||
|
||||
// MARK: - 全局音频上下文（C 回调无法访问 Swift 对象，用全局变量）
|
||||
private nonisolated(unsafe) var gAudioContext: AudioCaptureContext?
|
||||
// MARK: - 全局音频 Tap 上下文（C 回调无法访问 Swift 对象）
|
||||
|
||||
class AudioCaptureContext {
|
||||
var writerInput: AVAssetWriterInput?
|
||||
nonisolated(unsafe) var isRunning: Bool = true
|
||||
private nonisolated(unsafe) var gTapContext: AudioTapContext?
|
||||
|
||||
class AudioTapContext {
|
||||
nonisolated(unsafe) var writerInput: AVAssetWriterInput?
|
||||
nonisolated(unsafe) var isRecording: Bool = false
|
||||
nonisolated(unsafe) var formatDescription: CMAudioFormatDescription?
|
||||
nonisolated(unsafe) var sampleRate: Double = 44100.0
|
||||
nonisolated(unsafe) var channelsPerFrame: UInt32 = 2
|
||||
nonisolated(unsafe) var sampleRate: Double = 0
|
||||
nonisolated(unsafe) var channelsPerFrame: UInt32 = 0
|
||||
nonisolated(unsafe) var appendCount: Int = 0
|
||||
nonisolated(unsafe) var firstAppendLogged: Bool = false
|
||||
nonisolated(unsafe) var processCallCount: Int = 0
|
||||
nonisolated(unsafe) var firstAudioSampleTime: Int64 = -1
|
||||
nonisolated(unsafe) var firstAppendLogged: Bool = false
|
||||
nonisolated(unsafe) var totalFramesWritten: Int64 = 0
|
||||
|
||||
let audioQueue = DispatchQueue(label: "miniplayer.audioCapture")
|
||||
var dataBuffer: UnsafeMutableRawPointer?
|
||||
var dataBufferSize: UInt32 = 0
|
||||
/// 双缓冲区（消除数据竞争：写A时async读B，交替使用）
|
||||
nonisolated(unsafe) var dataBuffers: (UnsafeMutableRawPointer?, UnsafeMutableRawPointer?) = (nil, nil)
|
||||
nonisolated(unsafe) var dataBufferSizes: (Int, Int) = (0, 0)
|
||||
nonisolated(unsafe) var writeIndex: Int = 0
|
||||
|
||||
let audioQueue = DispatchQueue(label: "miniplayer.audioTap")
|
||||
}
|
||||
|
||||
// MARK: - CoreAudio 输出设备 IOProc 回调
|
||||
// MARK: - MTAudioProcessingTap C 回调
|
||||
|
||||
/// 从系统音频输出设备捕获正在播放的 PCM 数据
|
||||
/// 关键：输出设备的播放数据在 inOutputData 参数中，不是 inInputData
|
||||
private func audioDeviceIOProc(
|
||||
_ inDevice: AudioDeviceID,
|
||||
_ inNow: UnsafePointer<AudioTimeStamp>,
|
||||
_ inInputData: UnsafePointer<AudioBufferList>,
|
||||
_ inInputTime: UnsafePointer<AudioTimeStamp>,
|
||||
_ inOutputData: UnsafeMutablePointer<AudioBufferList>,
|
||||
_ inOutputTime: UnsafePointer<AudioTimeStamp>,
|
||||
_ inClientData: UnsafeMutableRawPointer?
|
||||
) -> OSStatus {
|
||||
guard let clientData = inClientData else { return noErr }
|
||||
let ctx = Unmanaged<AudioCaptureContext>.fromOpaque(clientData).takeUnretainedValue()
|
||||
guard ctx.isRunning else { return noErr }
|
||||
private func tapInit(
|
||||
_ tap: MTAudioProcessingTap,
|
||||
_ clientInfo: UnsafeMutableRawPointer?,
|
||||
_ tapStorageOut: UnsafeMutablePointer<UnsafeMutableRawPointer?>
|
||||
) {
|
||||
tapStorageOut.pointee = clientInfo
|
||||
}
|
||||
|
||||
private func tapFinalize(_ tap: MTAudioProcessingTap) {
|
||||
// Context lifecycle managed elsewhere
|
||||
}
|
||||
|
||||
private func tapPrepare(
|
||||
_ tap: MTAudioProcessingTap,
|
||||
_ maxFrames: CMItemCount,
|
||||
_ processingFormat: UnsafePointer<AudioStreamBasicDescription>
|
||||
) {
|
||||
let asbd = processingFormat.pointee
|
||||
NSLog("[Recorder] Tap prepare: rate=%.0f ch=%u formatFlags=0x%x",
|
||||
asbd.mSampleRate, asbd.mChannelsPerFrame, asbd.mFormatFlags)
|
||||
|
||||
let storage = MTAudioProcessingTapGetStorage(tap)
|
||||
let ctx = Unmanaged<AudioTapContext>.fromOpaque(storage).takeUnretainedValue()
|
||||
ctx.sampleRate = asbd.mSampleRate
|
||||
ctx.channelsPerFrame = asbd.mChannelsPerFrame
|
||||
|
||||
var fmtDesc: CMAudioFormatDescription?
|
||||
var mutableASBD = asbd
|
||||
CMAudioFormatDescriptionCreate(
|
||||
allocator: kCFAllocatorDefault,
|
||||
asbd: &mutableASBD,
|
||||
layoutSize: 0, layout: nil,
|
||||
magicCookieSize: 0, magicCookie: nil,
|
||||
extensions: nil,
|
||||
formatDescriptionOut: &fmtDesc
|
||||
)
|
||||
ctx.formatDescription = fmtDesc
|
||||
NSLog("[Recorder] Tap format ready: %@", fmtDesc != nil ? "OK" : "FAILED")
|
||||
}
|
||||
|
||||
private func tapUnprepare(_ tap: MTAudioProcessingTap) {
|
||||
NSLog("[Recorder] Tap unprepare")
|
||||
}
|
||||
|
||||
private func tapProcess(
|
||||
_ tap: MTAudioProcessingTap,
|
||||
_ numberFrames: Int,
|
||||
_ flags: UInt32,
|
||||
_ bufferListInOut: UnsafeMutablePointer<AudioBufferList>,
|
||||
_ bufferListSizeOut: UnsafeMutablePointer<Int>,
|
||||
_ flagsOut: UnsafeMutablePointer<UInt32>
|
||||
) {
|
||||
bufferListSizeOut.pointee = 1
|
||||
flagsOut.pointee = flags
|
||||
|
||||
let storage = MTAudioProcessingTapGetStorage(tap)
|
||||
let ctx = Unmanaged<AudioTapContext>.fromOpaque(storage).takeUnretainedValue()
|
||||
|
||||
ctx.processCallCount += 1
|
||||
guard ctx.isRecording,
|
||||
let _ = ctx.writerInput,
|
||||
let _ = ctx.formatDescription else { return }
|
||||
|
||||
// 从 inOutputData 读取正在播放的音频
|
||||
let outBufferList = inOutputData.pointee
|
||||
let inBufferList = inInputData.pointee
|
||||
let callNum = ctx.processCallCount
|
||||
if callNum <= 3 {
|
||||
print("[Recorder] IOProc#\(callNum): outBufs=\(outBufferList.mNumberBuffers), inBufs=\(inBufferList.mNumberBuffers)")
|
||||
if outBufferList.mNumberBuffers > 0 {
|
||||
let b = outBufferList.mBuffers
|
||||
print("[Recorder] outBuf[0]: size=\(b.mDataByteSize), channels=\(b.mNumberChannels), data=\(b.mData != nil)")
|
||||
}
|
||||
if inBufferList.mNumberBuffers > 0 {
|
||||
let b = inBufferList.mBuffers
|
||||
print("[Recorder] inBuf[0]: size=\(b.mDataByteSize), channels=\(b.mNumberChannels), data=\(b.mData != nil)")
|
||||
let ablPtr = UnsafeMutableAudioBufferListPointer(bufferListInOut)
|
||||
|
||||
// 计算总数据量
|
||||
var totalSize: Int = 0
|
||||
for i in 0..<ablPtr.count {
|
||||
if ablPtr[i].mData != nil && ablPtr[i].mDataByteSize > 0 {
|
||||
totalSize += Int(ablPtr[i].mDataByteSize)
|
||||
}
|
||||
}
|
||||
guard totalSize > 0 else { return }
|
||||
|
||||
// 优先读 inOutputData（系统正在播放的音频），这是 loopback 数据
|
||||
let bufferList: AudioBufferList
|
||||
if outBufferList.mNumberBuffers > 0 {
|
||||
let b = outBufferList.mBuffers
|
||||
if b.mDataByteSize > 0 && b.mData != nil {
|
||||
bufferList = outBufferList
|
||||
} else if inBufferList.mNumberBuffers > 0 {
|
||||
bufferList = inBufferList
|
||||
// 双缓冲：写当前槽，async读另一个槽（消除数据竞争）
|
||||
let wi = ctx.writeIndex
|
||||
ctx.writeIndex = 1 - wi // 切换槽位
|
||||
|
||||
if ctx.dataBufferSizes.0 < totalSize || ctx.dataBufferSizes.1 < totalSize {
|
||||
// 只在需要时扩容（不会在实时路径频繁触发）
|
||||
let newSize = max(totalSize, 65536) // 至少64KB
|
||||
if wi == 0 {
|
||||
if let existing = ctx.dataBuffers.0 { free(existing) }
|
||||
ctx.dataBuffers.0 = malloc(newSize)
|
||||
ctx.dataBufferSizes.0 = newSize
|
||||
} else {
|
||||
return noErr
|
||||
if let existing = ctx.dataBuffers.1 { free(existing) }
|
||||
ctx.dataBuffers.1 = malloc(newSize)
|
||||
ctx.dataBufferSizes.1 = newSize
|
||||
}
|
||||
} else if inBufferList.mNumberBuffers > 0 {
|
||||
bufferList = inBufferList
|
||||
} else {
|
||||
return noErr
|
||||
}
|
||||
|
||||
let buffer = bufferList.mBuffers
|
||||
let dataSize = buffer.mDataByteSize
|
||||
guard dataSize > 0, let srcData = buffer.mData else {
|
||||
if callNum <= 3 { print("[Recorder] IOProc#\(callNum): no data (size=\(buffer.mDataByteSize), data=\(buffer.mData != nil))") }
|
||||
return noErr
|
||||
let dest: UnsafeMutableRawPointer?
|
||||
if wi == 0 { dest = ctx.dataBuffers.0 } else { dest = ctx.dataBuffers.1 }
|
||||
guard let dest = dest else { return }
|
||||
|
||||
var offset = 0
|
||||
for i in 0..<ablPtr.count {
|
||||
if let src = ablPtr[i].mData, ablPtr[i].mDataByteSize > 0 {
|
||||
let sz = Int(ablPtr[i].mDataByteSize)
|
||||
memcpy(dest + offset, src, sz)
|
||||
offset += sz
|
||||
}
|
||||
}
|
||||
|
||||
// 复制到预分配 buffer
|
||||
if ctx.dataBufferSize < dataSize {
|
||||
if let existing = ctx.dataBuffer { free(existing) }
|
||||
ctx.dataBuffer = malloc(Int(dataSize))
|
||||
ctx.dataBufferSize = dataSize
|
||||
}
|
||||
guard let dest = ctx.dataBuffer else { return noErr }
|
||||
memcpy(dest, srcData, Int(dataSize))
|
||||
let frameCount = Int64(numberFrames)
|
||||
let startFrame = ctx.totalFramesWritten
|
||||
let readWi = wi // async 读取的是刚写完的槽
|
||||
|
||||
let pts = inOutputTime.pointee
|
||||
|
||||
// 在串行队列处理
|
||||
// 在串行队列创建 CMSampleBuffer 并写入
|
||||
ctx.audioQueue.async {
|
||||
guard ctx.isRunning else { return }
|
||||
guard let fd = ctx.formatDescription, let writerInput = ctx.writerInput else { return }
|
||||
guard ctx.isRecording,
|
||||
let writerInput = ctx.writerInput,
|
||||
let fd = ctx.formatDescription else { return }
|
||||
|
||||
// 从已写完的槽读取（此时实时线程正在写另一个槽，安全）
|
||||
let readDest: UnsafeMutableRawPointer?
|
||||
if readWi == 0 { readDest = ctx.dataBuffers.0 } else { readDest = ctx.dataBuffers.1 }
|
||||
guard let readDest = readDest else { return }
|
||||
|
||||
var blockBuffer: CMBlockBuffer?
|
||||
let blockStatus = CMBlockBufferCreateWithMemoryBlock(
|
||||
allocator: kCFAllocatorDefault, memoryBlock: nil,
|
||||
blockLength: Int(dataSize), blockAllocator: kCFAllocatorDefault,
|
||||
blockLength: totalSize, blockAllocator: kCFAllocatorDefault,
|
||||
customBlockSource: nil, offsetToData: 0,
|
||||
dataLength: Int(dataSize),
|
||||
dataLength: totalSize,
|
||||
flags: kCMBlockBufferAssureMemoryNowFlag,
|
||||
blockBufferOut: &blockBuffer
|
||||
)
|
||||
guard blockStatus == kCMBlockBufferNoErr, let bb = blockBuffer else {
|
||||
if callNum <= 3 { print("[Recorder] CMBlockBufferCreate failed: \(blockStatus)") }
|
||||
return
|
||||
}
|
||||
guard blockStatus == kCMBlockBufferNoErr, let bb = blockBuffer else { return }
|
||||
|
||||
let replaceStatus = CMBlockBufferReplaceDataBytes(
|
||||
with: dest, blockBuffer: bb,
|
||||
offsetIntoDestination: 0, dataLength: Int(dataSize)
|
||||
with: readDest, blockBuffer: bb,
|
||||
offsetIntoDestination: 0, dataLength: totalSize
|
||||
)
|
||||
guard replaceStatus == kCMBlockBufferNoErr else { return }
|
||||
|
||||
var timingInfo = CMSampleTimingInfo()
|
||||
if pts.mFlags.contains(.sampleTimeValid) {
|
||||
// 使用相对于第一次音频回调的时间戳
|
||||
if ctx.firstAudioSampleTime < 0 {
|
||||
ctx.firstAudioSampleTime = Int64(pts.mSampleTime)
|
||||
}
|
||||
let relativeSampleTime = Int64(pts.mSampleTime) - ctx.firstAudioSampleTime
|
||||
timingInfo.presentationTimeStamp = CMTime(
|
||||
value: relativeSampleTime,
|
||||
timescale: Int32(ctx.sampleRate)
|
||||
)
|
||||
} else if pts.mFlags.contains(.hostTimeValid) {
|
||||
var timebase = mach_timebase_info_data_t()
|
||||
mach_timebase_info(&timebase)
|
||||
let nanos = pts.mHostTime * UInt64(timebase.numer) / UInt64(timebase.denom)
|
||||
timingInfo.presentationTimeStamp = CMTime(value: CMTimeValue(nanos), timescale: 1_000_000_000)
|
||||
} else {
|
||||
return
|
||||
}
|
||||
timingInfo.duration = .invalid
|
||||
timingInfo.decodeTimeStamp = .invalid
|
||||
|
||||
let bytesPerFrame = Int(ctx.channelsPerFrame) * MemoryLayout<Float32>.size
|
||||
let sampleCount = Int(dataSize) / max(bytesPerFrame, 1)
|
||||
// PTS 基于累计帧数计算
|
||||
let pts = CMTime(
|
||||
value: CMTimeValue(startFrame),
|
||||
timescale: Int32(ctx.sampleRate)
|
||||
)
|
||||
var timingInfo = CMSampleTimingInfo(
|
||||
duration: CMTime(value: CMTimeValue(frameCount), timescale: Int32(ctx.sampleRate)),
|
||||
presentationTimeStamp: pts,
|
||||
decodeTimeStamp: .invalid
|
||||
)
|
||||
|
||||
var sampleBuffer: CMSampleBuffer?
|
||||
let createStatus = CMSampleBufferCreateReady(
|
||||
allocator: kCFAllocatorDefault,
|
||||
dataBuffer: bb,
|
||||
formatDescription: fd,
|
||||
sampleCount: sampleCount,
|
||||
sampleCount: Int(frameCount),
|
||||
sampleTimingEntryCount: 1,
|
||||
sampleTimingArray: &timingInfo,
|
||||
sampleSizeEntryCount: 0,
|
||||
@ -157,40 +191,33 @@ private func audioDeviceIOProc(
|
||||
sampleBufferOut: &sampleBuffer
|
||||
)
|
||||
guard createStatus == noErr, let sb = sampleBuffer else {
|
||||
if callNum <= 3 {
|
||||
print("[Recorder] ⚠️ SampleBuffer create failed: \(createStatus), dataSize=\(dataSize), sampleCount=\(sampleCount)")
|
||||
}
|
||||
NSLog("[Recorder] ⚠️ SampleBuffer create failed: %d", createStatus)
|
||||
return
|
||||
}
|
||||
|
||||
if writerInput.isReadyForMoreMediaData {
|
||||
let appended = writerInput.append(sb)
|
||||
if appended {
|
||||
if writerInput.append(sb) {
|
||||
ctx.appendCount += 1
|
||||
ctx.totalFramesWritten += frameCount
|
||||
if !ctx.firstAppendLogged {
|
||||
ctx.firstAppendLogged = true
|
||||
print("[Recorder] ✓ First audio captured: pts=\(timingInfo.presentationTimeStamp.seconds)s, size=\(dataSize)")
|
||||
NSLog("[Recorder] ✓ First audio captured: pts=%.3fs, frames=%lld, size=%d",
|
||||
pts.seconds, frameCount, totalSize)
|
||||
}
|
||||
} else if callNum <= 3 {
|
||||
print("[Recorder] ⚠️ writerInput.append returned false")
|
||||
}
|
||||
} else if callNum <= 3 {
|
||||
print("[Recorder] ⚠️ writerInput not ready for more data")
|
||||
}
|
||||
}
|
||||
|
||||
return noErr
|
||||
}
|
||||
|
||||
// MARK: - PlayerRecorder
|
||||
// MARK: - RecorderState
|
||||
|
||||
/// 全局状态：录制是否正在 finalize（阻止应用退出）
|
||||
enum RecorderState {
|
||||
static nonisolated(unsafe) var isFinalizing: Bool = false
|
||||
static nonisolated(unsafe) var pendingTerminate: Bool = false
|
||||
}
|
||||
|
||||
/// 视频源录制器：AVPlayerItemVideoOutput 抓帧 + CoreAudio 输出设备捕获音频
|
||||
// MARK: - PlayerRecorder
|
||||
|
||||
@MainActor
|
||||
final class PlayerRecorder: NSObject {
|
||||
|
||||
@ -206,10 +233,8 @@ final class PlayerRecorder: NSObject {
|
||||
private var lastPixelBuffer: CVPixelBuffer?
|
||||
private var captureFrameCount: Int = 0
|
||||
|
||||
// 音频 (CoreAudio)
|
||||
private var audioDeviceID: AudioDeviceID = 0
|
||||
private var audioCaptureContext: AudioCaptureContext?
|
||||
private var audioIOProc: AudioDeviceIOProc?
|
||||
// Audio Tap
|
||||
private var tapContext: AudioTapContext?
|
||||
private weak var currentPlayerItem: AVPlayerItem?
|
||||
|
||||
nonisolated(unsafe) private var recordStartTime: CMTime = .zero
|
||||
@ -229,6 +254,61 @@ final class PlayerRecorder: NSObject {
|
||||
return dir
|
||||
}
|
||||
|
||||
// MARK: - Audio Tap Setup
|
||||
|
||||
/// 在录制开始时安装音频 Tap 到 playerItem
|
||||
private func installAudioTap(on playerItem: AVPlayerItem) {
|
||||
// 确保全局上下文存在
|
||||
if gTapContext == nil {
|
||||
gTapContext = AudioTapContext()
|
||||
}
|
||||
guard let ctx = gTapContext else { return }
|
||||
|
||||
// 同步获取音频 track（录制时 asset 已经加载完毕）
|
||||
let audioTracks = playerItem.asset.tracks(withMediaType: .audio)
|
||||
guard let audioTrack = audioTracks.first else {
|
||||
NSLog("[Recorder] ⚠️ No audio track found (tried %d tracks)", audioTracks.count)
|
||||
return
|
||||
}
|
||||
|
||||
NSLog("[Recorder] Found audio track: ID=%d, formatDescriptions=%d",
|
||||
audioTrack.trackID, audioTrack.formatDescriptions.count)
|
||||
|
||||
// 创建 MTAudioProcessingTap
|
||||
var callbacks = MTAudioProcessingTapCallbacks(
|
||||
version: kMTAudioProcessingTapCallbacksVersion_0,
|
||||
clientInfo: Unmanaged.passUnretained(ctx).toOpaque(),
|
||||
init: tapInit,
|
||||
finalize: tapFinalize,
|
||||
prepare: tapPrepare,
|
||||
unprepare: tapUnprepare,
|
||||
process: tapProcess
|
||||
)
|
||||
|
||||
var tap: MTAudioProcessingTap?
|
||||
let status = MTAudioProcessingTapCreate(
|
||||
kCFAllocatorDefault, &callbacks,
|
||||
kMTAudioProcessingTapCreationFlag_PreEffects,
|
||||
&tap
|
||||
)
|
||||
guard status == noErr, let tapRef = tap else {
|
||||
NSLog("[Recorder] ⚠️ MTAudioProcessingTapCreate failed: %d", status)
|
||||
return
|
||||
}
|
||||
|
||||
// 创建 AudioMix — 用 kCMPersistentTrackID_Invalid 匹配所有音频轨道
|
||||
// HLS 流在 variant 切换时 audio track ID 会变，绑定具体 ID 会导致 tap 失效
|
||||
let inputParams = AVMutableAudioMixInputParameters()
|
||||
inputParams.trackID = kCMPersistentTrackID_Invalid
|
||||
inputParams.audioTapProcessor = tapRef
|
||||
|
||||
let audioMix = AVMutableAudioMix()
|
||||
audioMix.inputParameters = [inputParams]
|
||||
|
||||
playerItem.audioMix = audioMix
|
||||
NSLog("[Recorder] ✓ Audio tap installed on playerItem, trackID=%d", audioTrack.trackID)
|
||||
}
|
||||
|
||||
// MARK: - 开始录制
|
||||
|
||||
func startRecording(from output: AVPlayerItemVideoOutput, playerItem: AVPlayerItem, startTime: CMTime) {
|
||||
@ -241,11 +321,19 @@ final class PlayerRecorder: NSObject {
|
||||
isRunning = false
|
||||
recordStartTime = startTime
|
||||
|
||||
// 安装音频 Tap（录制时 asset 已加载，同步获取 track）
|
||||
installAudioTap(on: playerItem)
|
||||
|
||||
let tempFilename = "temp_recording_\(Int(Date().timeIntervalSince1970)).mp4"
|
||||
let url = saveDirectory.appendingPathComponent(tempFilename)
|
||||
try? FileManager.default.removeItem(at: url)
|
||||
tempURL = url
|
||||
|
||||
// 获取 tap context（可能还没有 format，先用默认值）
|
||||
let ctx = gTapContext ?? AudioTapContext()
|
||||
if gTapContext == nil { gTapContext = ctx }
|
||||
self.tapContext = ctx
|
||||
|
||||
do {
|
||||
writer = try AVAssetWriter(outputURL: url, fileType: .mp4)
|
||||
} catch {
|
||||
@ -268,7 +356,50 @@ final class PlayerRecorder: NSObject {
|
||||
videoInput = vInput
|
||||
}
|
||||
|
||||
// 音频输入将在 startAudioCapture 中获取设备格式后用 AAC 编码创建
|
||||
// 音频输入（AAC 编码）— 在 startWriting 之前添加
|
||||
// 使用 tap 已知的格式，或默认值
|
||||
let audioRate = ctx.sampleRate > 0 ? ctx.sampleRate : 48000.0
|
||||
let audioCh = ctx.channelsPerFrame > 0 ? ctx.channelsPerFrame : 2
|
||||
|
||||
var sourceFmtDesc: CMAudioFormatDescription?
|
||||
if let tapFmt = ctx.formatDescription {
|
||||
sourceFmtDesc = tapFmt
|
||||
} else {
|
||||
// Tap 还没有 prepare，创建默认格式描述
|
||||
var asbd = AudioStreamBasicDescription(
|
||||
mSampleRate: audioRate,
|
||||
mFormatID: kAudioFormatLinearPCM,
|
||||
mFormatFlags: kAudioFormatFlagIsFloat | kAudioFormatFlagIsPacked | kAudioFormatFlagIsNonInterleaved,
|
||||
mBytesPerPacket: 4,
|
||||
mFramesPerPacket: 1,
|
||||
mBytesPerFrame: 4,
|
||||
mChannelsPerFrame: audioCh,
|
||||
mBitsPerChannel: 32,
|
||||
mReserved: 0
|
||||
)
|
||||
CMAudioFormatDescriptionCreate(
|
||||
allocator: kCFAllocatorDefault,
|
||||
asbd: &asbd, layoutSize: 0, layout: nil,
|
||||
magicCookieSize: 0, magicCookie: nil,
|
||||
extensions: nil, formatDescriptionOut: &sourceFmtDesc
|
||||
)
|
||||
}
|
||||
|
||||
if let fmtDesc = sourceFmtDesc {
|
||||
let audioSettings: [String: Any] = [
|
||||
AVFormatIDKey: kAudioFormatMPEG4AAC,
|
||||
AVSampleRateKey: audioRate,
|
||||
AVNumberOfChannelsKey: audioCh,
|
||||
AVEncoderBitRateKey: 128_000
|
||||
]
|
||||
let aInput = AVAssetWriterInput(mediaType: .audio, outputSettings: audioSettings, sourceFormatHint: fmtDesc)
|
||||
aInput.expectsMediaDataInRealTime = true
|
||||
if writer.canAdd(aInput) {
|
||||
writer.add(aInput)
|
||||
audioInput = aInput
|
||||
NSLog("[Recorder] AudioInput added: AAC rate=%.0f ch=%u", audioRate, audioCh)
|
||||
}
|
||||
}
|
||||
|
||||
guard writer.startWriting() else {
|
||||
NSLog("[Recorder] startWriting failed: %@", writer.error?.localizedDescription ?? "unknown")
|
||||
@ -279,13 +410,18 @@ final class PlayerRecorder: NSObject {
|
||||
writer.startSession(atSourceTime: .zero)
|
||||
NSLog("[Recorder] Writer started, writing to: %@", url.lastPathComponent)
|
||||
|
||||
// 重置 tap context 录制状态
|
||||
ctx.isRecording = true
|
||||
ctx.writerInput = audioInput
|
||||
ctx.appendCount = 0
|
||||
ctx.totalFramesWritten = 0
|
||||
ctx.firstAppendLogged = false
|
||||
ctx.processCallCount = 0
|
||||
|
||||
isRecording = true
|
||||
isRunning = true
|
||||
startCaptureLoop()
|
||||
startDurationTimer()
|
||||
|
||||
// 启动音频捕获 (CoreAudio 输出设备，内部创建 audioInput)
|
||||
startAudioCapture()
|
||||
}
|
||||
|
||||
// MARK: - 停止录制
|
||||
@ -299,11 +435,14 @@ final class PlayerRecorder: NSObject {
|
||||
captureTimer = nil
|
||||
stopDurationTimer()
|
||||
|
||||
let audioCalls = audioCaptureContext?.processCallCount ?? 0
|
||||
let audioAppended = audioCaptureContext?.appendCount ?? 0
|
||||
// 停止音频 tap 写入
|
||||
tapContext?.isRecording = false
|
||||
let audioAppended = tapContext?.appendCount ?? 0
|
||||
let audioCalls = tapContext?.processCallCount ?? 0
|
||||
|
||||
NSLog("[Recorder] Stopping: videoFrames=%d, audioProcessCalls=%d, audioAppended=%d",
|
||||
captureFrameCount, audioCalls, audioAppended)
|
||||
stopAudioCapture()
|
||||
|
||||
currentPlayerItem = nil
|
||||
|
||||
videoInput?.markAsFinished()
|
||||
@ -352,121 +491,6 @@ final class PlayerRecorder: NSObject {
|
||||
|
||||
lastPixelBuffer = nil
|
||||
}
|
||||
|
||||
// MARK: - 音频捕获 (CoreAudio 输出设备)
|
||||
|
||||
private func startAudioCapture() {
|
||||
print("[Recorder] startAudioCapture called")
|
||||
|
||||
var deviceID: AudioDeviceID = 0
|
||||
var size = UInt32(MemoryLayout<AudioDeviceID>.size)
|
||||
var address = AudioObjectPropertyAddress(
|
||||
mSelector: kAudioHardwarePropertyDefaultOutputDevice,
|
||||
mScope: kAudioObjectPropertyScopeGlobal,
|
||||
mElement: kAudioObjectPropertyElementMain
|
||||
)
|
||||
let status = AudioObjectGetPropertyData(
|
||||
AudioObjectID(kAudioObjectSystemObject), &address, 0, nil, &size, &deviceID
|
||||
)
|
||||
print("[Recorder] Get default device status: \(status), deviceID: \(deviceID)")
|
||||
guard status == noErr, deviceID != 0 else {
|
||||
print("[Recorder] ❌ Cannot get default audio output device")
|
||||
return
|
||||
}
|
||||
self.audioDeviceID = deviceID
|
||||
|
||||
// 获取输出音频流格式
|
||||
var streamFormatAddress = AudioObjectPropertyAddress(
|
||||
mSelector: kAudioDevicePropertyStreamFormat,
|
||||
mScope: kAudioDevicePropertyScopeOutput,
|
||||
mElement: kAudioObjectPropertyElementMain
|
||||
)
|
||||
var asbd = AudioStreamBasicDescription()
|
||||
size = UInt32(MemoryLayout<AudioStreamBasicDescription>.size)
|
||||
let fmtStatus = AudioObjectGetPropertyData(deviceID, &streamFormatAddress, 0, nil, &size, &asbd)
|
||||
print("[Recorder] Stream format: status=\(fmtStatus), rate=\(asbd.mSampleRate), ch=\(asbd.mChannelsPerFrame), bits=\(asbd.mBitsPerChannel), formatID=\(asbd.mFormatID)")
|
||||
guard fmtStatus == noErr else {
|
||||
print("[Recorder] ❌ Cannot get audio stream format")
|
||||
return
|
||||
}
|
||||
|
||||
// 用 AAC 编码创建音频输入（MP4 容器不支持裸 PCM pass-through）
|
||||
// 先创建 CMAudioFormatDescription
|
||||
var fmtDesc: CMAudioFormatDescription?
|
||||
var mutableASBD = asbd
|
||||
CMAudioFormatDescriptionCreate(
|
||||
allocator: kCFAllocatorDefault,
|
||||
asbd: &mutableASBD,
|
||||
layoutSize: 0, layout: nil,
|
||||
magicCookieSize: 0, magicCookie: nil,
|
||||
extensions: nil,
|
||||
formatDescriptionOut: &fmtDesc
|
||||
)
|
||||
guard let sourceFmtDesc = fmtDesc else {
|
||||
print("[Recorder] ❌ Cannot create CMAudioFormatDescription")
|
||||
return
|
||||
}
|
||||
|
||||
let audioSettings: [String: Any] = [
|
||||
AVFormatIDKey: kAudioFormatMPEG4AAC,
|
||||
AVSampleRateKey: asbd.mSampleRate,
|
||||
AVNumberOfChannelsKey: asbd.mChannelsPerFrame,
|
||||
AVEncoderBitRateKey: 128_000
|
||||
]
|
||||
let aInput = AVAssetWriterInput(mediaType: .audio, outputSettings: audioSettings, sourceFormatHint: sourceFmtDesc)
|
||||
aInput.expectsMediaDataInRealTime = true
|
||||
if let w = writer, w.canAdd(aInput) {
|
||||
w.add(aInput)
|
||||
audioInput = aInput
|
||||
print("[Recorder] ✓ AudioInput created with AAC encoding (rate=\(asbd.mSampleRate), ch=\(asbd.mChannelsPerFrame))")
|
||||
} else {
|
||||
print("[Recorder] ❌ Cannot add audio input to writer")
|
||||
return
|
||||
}
|
||||
|
||||
// 创建 context
|
||||
let context = AudioCaptureContext()
|
||||
context.writerInput = aInput
|
||||
context.sampleRate = Double(asbd.mSampleRate)
|
||||
context.channelsPerFrame = asbd.mChannelsPerFrame
|
||||
context.isRunning = true
|
||||
self.audioCaptureContext = context
|
||||
|
||||
// 使用前面已创建的格式描述
|
||||
context.formatDescription = sourceFmtDesc
|
||||
print("[Recorder] formatDescription created: true")
|
||||
|
||||
// 创建 IOProc
|
||||
let contextPtr = Unmanaged.passUnretained(context).toOpaque()
|
||||
var ioProc: AudioDeviceIOProc?
|
||||
let createStatus = AudioDeviceCreateIOProcID(deviceID, audioDeviceIOProc, contextPtr, &ioProc)
|
||||
print("[Recorder] CreateIOProcID status: \(createStatus)")
|
||||
guard createStatus == noErr, let proc = ioProc else {
|
||||
print("[Recorder] ❌ AudioDeviceCreateIOProcID failed: \(createStatus)")
|
||||
return
|
||||
}
|
||||
self.audioIOProc = proc
|
||||
|
||||
let startStatus = AudioDeviceStart(deviceID, proc)
|
||||
print("[Recorder] AudioDeviceStart status: \(startStatus)")
|
||||
guard startStatus == noErr else {
|
||||
print("[Recorder] ❌ AudioDeviceStart failed: \(startStatus)")
|
||||
AudioDeviceDestroyIOProcID(deviceID, proc)
|
||||
self.audioIOProc = nil
|
||||
return
|
||||
}
|
||||
NSLog("[Recorder] ✓ CoreAudio output device capture started (deviceID=%u)", deviceID)
|
||||
}
|
||||
|
||||
private func stopAudioCapture() {
|
||||
audioCaptureContext?.isRunning = false
|
||||
if let proc = audioIOProc, audioDeviceID != 0 {
|
||||
AudioDeviceStop(audioDeviceID, proc)
|
||||
AudioDeviceDestroyIOProcID(audioDeviceID, proc)
|
||||
}
|
||||
audioIOProc = nil
|
||||
audioCaptureContext = nil
|
||||
}
|
||||
|
||||
// MARK: - 视频抓帧 (30fps Timer)
|
||||
|
||||
|
||||
Loading…
x
Reference in New Issue
Block a user