MiniPlayer/Sources/PlayerRecorder.swift
yumoqing 339664ebc1 fix: 修复HLS流媒体录制无音频问题
- 存储tapPrepareCallback实际采样率,替代不可靠的timeRange.duration.timescale
- findAudioTrackID增加3路fallback: playerItem.tracks → asset.tracks → 非视频推断
- async load(.tracks)从单次尝试改为每3秒重试
- 增加muxed track fallback(HLS可能返回非标准mediaType)
2026-06-26 08:51:10 +08:00

721 lines
29 KiB
Swift
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

import AVFoundation
import CoreMedia
import MediaToolbox
import Foundation
#if os(macOS)
import AppKit
// MARK: - C 访 Swift
private nonisolated(unsafe) var gAudioContext: AudioCaptureContext?
class AudioCaptureContext {
var writerInput: AVAssetWriterInput?
nonisolated(unsafe) var isRunning: Bool = true
nonisolated(unsafe) var formatDescription: CMAudioFormatDescription?
nonisolated(unsafe) var appendCount: Int = 0
nonisolated(unsafe) var firstAppendLogged: Bool = false
nonisolated(unsafe) var processCallCount: Int = 0
nonisolated(unsafe) var sampleRate: Double = 44100.0
nonisolated(unsafe) var channelsPerFrame: UInt32 = 2
}
// MARK: - MTAudioProcessingTap C
private func tapInitCallback(
_ tap: MTAudioProcessingTap,
_ clientInfo: UnsafeMutableRawPointer?,
_ tapStorageOut: UnsafeMutablePointer<UnsafeMutableRawPointer?>
) {
// tapStorage context 访
tapStorageOut.pointee = clientInfo
}
private func tapFinalizeCallback(_ tap: MTAudioProcessingTap) {
// stopRecording
}
private func tapPrepareCallback(
_ tap: MTAudioProcessingTap,
_ maxFrames: CMItemCount,
_ processingFormat: UnsafePointer<AudioStreamBasicDescription>
) {
guard let ctx = gAudioContext else { return }
var asbd = processingFormat.pointee
ctx.sampleRate = Double(asbd.mSampleRate)
ctx.channelsPerFrame = asbd.mChannelsPerFrame
var fmtDesc: CMAudioFormatDescription?
CMAudioFormatDescriptionCreate(
allocator: kCFAllocatorDefault,
asbd: &asbd,
layoutSize: 0, layout: nil,
magicCookieSize: 0, magicCookie: nil,
extensions: nil,
formatDescriptionOut: &fmtDesc
)
ctx.formatDescription = fmtDesc
NSLog("[Recorder] tapPrepare: %.1fHz, %uch, mFormatID=%u", asbd.mSampleRate, asbd.mChannelsPerFrame, asbd.mFormatID)
}
private func tapUnprepareCallback(_ tap: MTAudioProcessingTap) {
// Nothing to clean up
}
private func tapProcessCallback(
_ tap: MTAudioProcessingTap,
_ numberFrames: CMItemCount,
_ flags: MTAudioProcessingTapFlags,
_ bufferListInOut: UnsafeMutablePointer<AudioBufferList>,
_ numberFramesOut: UnsafeMutablePointer<CMItemCount>,
_ flagsOut: UnsafeMutablePointer<MTAudioProcessingTapFlags>
) {
//
var timeRange = CMTimeRange()
var srcFlags: MTAudioProcessingTapFlags = 0
var actualFrames: CMItemCount = 0
let status = MTAudioProcessingTapGetSourceAudio(
tap, numberFrames, bufferListInOut,
&srcFlags, &timeRange, &actualFrames
)
//
numberFramesOut.pointee = actualFrames
flagsOut.pointee = srcFlags
guard status == noErr, actualFrames > 0 else { return }
// context
guard let ctx = gAudioContext, ctx.isRunning,
let writerInput = ctx.writerInput,
writerInput.isReadyForMoreMediaData else { return }
ctx.processCallCount += 1
// 500
if ctx.processCallCount == 1 || ctx.processCallCount % 500 == 0 {
NSLog("[Recorder] tapProcess #%d: frames=%ld, timescale=%d",
ctx.processCallCount, actualFrames, timeRange.duration.timescale)
}
let bufferList = bufferListInOut.pointee
guard bufferList.mNumberBuffers > 0 else { return }
let buf = bufferList.mBuffers
let dataSize = Int(buf.mDataByteSize)
guard dataSize > 0, let srcData = buf.mData else { return }
// CMBlockBuffer
var blockBuffer: CMBlockBuffer?
let blockStatus = CMBlockBufferCreateWithMemoryBlock(
allocator: kCFAllocatorDefault, memoryBlock: nil,
blockLength: dataSize, blockAllocator: kCFAllocatorDefault,
customBlockSource: nil, offsetToData: 0,
dataLength: dataSize,
flags: kCMBlockBufferAssureMemoryNowFlag,
blockBufferOut: &blockBuffer
)
guard blockStatus == kCMBlockBufferNoErr, let bb = blockBuffer else { return }
let replaceStatus = CMBlockBufferReplaceDataBytes(
with: srcData, blockBuffer: bb,
offsetIntoDestination: 0, dataLength: dataSize
)
guard replaceStatus == kCMBlockBufferNoErr else { return }
// format description
guard let fd = ctx.formatDescription else {
if ctx.processCallCount == 1 {
NSLog("[Recorder] ⚠️ formatDescription nil on first process call")
}
return
}
// PTS 0
let sampleRate = ctx.sampleRate
let ptsValue = Double(ctx.appendCount) / sampleRate
let pts = CMTime(seconds: ptsValue, preferredTimescale: Int32(sampleRate))
var sampleBuffer: CMSampleBuffer?
let createStatus = CMAudioSampleBufferCreateReadyWithPacketDescriptions(
allocator: kCFAllocatorDefault,
dataBuffer: bb,
formatDescription: fd,
sampleCount: actualFrames,
presentationTimeStamp: pts,
packetDescriptions: nil,
sampleBufferOut: &sampleBuffer
)
guard createStatus == noErr, let sb = sampleBuffer else {
if ctx.processCallCount <= 3 {
NSLog("[Recorder] ⚠️ CMAudioSampleBufferCreate failed: %d", createStatus)
}
return
}
if writerInput.append(sb) {
ctx.appendCount += actualFrames
if !ctx.firstAppendLogged {
ctx.firstAppendLogged = true
NSLog("[Recorder] ✓ First audio appended: pts=%.2fs, frames=%ld", pts.seconds, actualFrames)
}
} else if ctx.processCallCount <= 3 {
NSLog("[Recorder] ⚠️ writerInput.append failed (input ready=%d)", writerInput.isReadyForMoreMediaData)
}
}
// MARK: - PlayerRecorder
/// finalize退
enum RecorderState {
static nonisolated(unsafe) var isFinalizing: Bool = false
static nonisolated(unsafe) var pendingTerminate: Bool = false
}
/// AVPlayerItemVideoOutput + MTAudioProcessingTap
///
@MainActor
final class PlayerRecorder: NSObject {
private var writer: AVAssetWriter?
private var videoInput: AVAssetWriterInput?
private var audioInput: AVAssetWriterInput?
private var captureTimer: Timer?
private var durationTimer: Timer?
private var startDate: Date?
private weak var weakOutput: AVPlayerItemVideoOutput?
nonisolated(unsafe) private var isRunning = false
//
private var lastPixelBuffer: CVPixelBuffer?
private var captureFrameCount: Int = 0
// (MTAudioProcessingTap)
private var audioTap: MTAudioProcessingTap?
private var audioContext: AudioCaptureContext?
private weak var currentPlayerItem: AVPlayerItem?
private var tracksObservation: NSKeyValueObservation?
//
nonisolated(unsafe) private var recordStartTime: CMTime = .zero
@Published var isRecording = false
@Published var durationText = "00:00"
var onRecordingSaved: ((URL) -> Void)?
var onError: ((String) -> Void)?
///
private var tempURL: URL?
///
private var saveDirectory: URL {
let movies = FileManager.default.urls(for: .moviesDirectory, in: .userDomainMask).first!
let dir = movies.appendingPathComponent("MiniPlayer", isDirectory: true)
try? FileManager.default.createDirectory(at: dir, withIntermediateDirectories: true)
return dir
}
// MARK: -
func startRecording(from output: AVPlayerItemVideoOutput, playerItem: AVPlayerItem, startTime: CMTime) {
guard !isRecording else { return }
weakOutput = output
currentPlayerItem = playerItem
lastPixelBuffer = nil
captureFrameCount = 0
isRunning = false
recordStartTime = startTime
//
let tempFilename = "temp_recording_\(Int(Date().timeIntervalSince1970)).mp4"
let url = saveDirectory.appendingPathComponent(tempFilename)
//
try? FileManager.default.removeItem(at: url)
tempURL = url
do {
writer = try AVAssetWriter(outputURL: url, fileType: .mp4)
} catch {
onError?("Cannot create writer: \(error.localizedDescription)")
return
}
guard let writer = writer else { return }
// H264
let videoSettings: [String: Any] = [
AVVideoCodecKey: AVVideoCodecType.h264,
AVVideoWidthKey: 1920,
AVVideoHeightKey: 1080
]
let vInput = AVAssetWriterInput(mediaType: .video, outputSettings: videoSettings)
vInput.expectsMediaDataInRealTime = true
if writer.canAdd(vInput) {
writer.add(vInput)
videoInput = vInput
}
// AAC
let audioSettings: [String: Any] = [
AVFormatIDKey: kAudioFormatMPEG4AAC,
AVSampleRateKey: 44100,
AVNumberOfChannelsKey: 2,
AVEncoderBitRateKey: 128000
]
let aInput = AVAssetWriterInput(mediaType: .audio, outputSettings: audioSettings)
aInput.expectsMediaDataInRealTime = true
if writer.canAdd(aInput) {
writer.add(aInput)
audioInput = aInput
}
guard writer.startWriting() else {
NSLog("[Recorder] startWriting failed: %@", writer.error?.localizedDescription ?? "unknown")
onError?("Cannot start writer: \(writer.error?.localizedDescription ?? "unknown")")
return
}
// session 0
writer.startSession(atSourceTime: .zero)
NSLog("[Recorder] Writer started, writing to: %@", url.lastPathComponent)
// 30fps
isRecording = true
isRunning = true
startCaptureLoop()
startDurationTimer()
// MTAudioProcessingTap
setupAudioTap(playerItem: playerItem, audioWriterInput: aInput)
}
// MARK: -
func stopRecording() {
guard isRecording else { return }
isRecording = false
isRunning = false
RecorderState.isFinalizing = true // 退 finishWriting
captureTimer?.invalidate()
captureTimer = nil
stopDurationTimer()
// tap
audioContext?.isRunning = false
tracksObservation = nil
let totalAudioFrames = audioContext?.appendCount ?? 0
let processCalls = audioContext?.processCallCount ?? 0
NSLog("[Recorder] Stopping: videoFrames=%d, audioProcessCalls=%d, audioAppended=%d",
captureFrameCount, processCalls, totalAudioFrames)
// audioMix
currentPlayerItem?.audioMix = nil
currentPlayerItem = nil
// tap context
audioTap = nil
audioContext = nil
gAudioContext = nil
videoInput?.markAsFinished()
audioInput?.markAsFinished()
writer?.finishWriting { [weak self] in
NSLog("[Recorder] finishWriting callback invoked")
DispatchQueue.main.async {
RecorderState.isFinalizing = false // 退
guard let self = self, let w = self.writer else {
NSLog("[Recorder] ⚠️ self or writer is nil in finishWriting callback")
if RecorderState.pendingTerminate {
RecorderState.pendingTerminate = false
NSApp.reply(toApplicationShouldTerminate: true)
}
return
}
let vFrames = self.captureFrameCount
NSLog("[Recorder] finishWriting: status=%d, error=%@, videoFrames=%d, audioFrames=%d",
w.status.rawValue,
w.error?.localizedDescription ?? "none",
vFrames, totalAudioFrames)
if w.status == .completed, let tempURL = self.tempURL {
//
if FileManager.default.fileExists(atPath: tempURL.path) {
let size = (try? FileManager.default.attributesOfItem(atPath: tempURL.path)[.size] as? UInt64) ?? 0
NSLog("[Recorder] ✓ Temp file exists: %@, size=%llu bytes", tempURL.lastPathComponent, size)
self.showSaveDialog(tempURL: tempURL)
} else {
NSLog("[Recorder] ⚠️ Temp file missing: %@", tempURL.path)
self.onError?("Recording file missing")
self.checkPendingTerminate()
}
} else {
NSLog("[Recorder] ✗ finishWriting failed: status=%d, error=%@",
w.status.rawValue, w.error?.localizedDescription ?? "unknown")
self.onError?("Recording failed: \(w.error?.localizedDescription ?? "unknown (status=\(w.status.rawValue))")")
if let url = self.tempURL {
try? FileManager.default.removeItem(at: url)
}
self.checkPendingTerminate()
}
}
}
lastPixelBuffer = nil
}
// MARK: - Tap
private func setupAudioTap(playerItem: AVPlayerItem, audioWriterInput: AVAssetWriterInput) {
// HLS tracks
// KVO async load
// track
let tracks = playerItem.tracks
NSLog("[Recorder] setupAudioTap: playerItem.tracks.count=%d", tracks.count)
for (i, track) in tracks.enumerated() {
if let at = track.assetTrack {
NSLog("[Recorder] track[%d]: mediaType=%@, trackID=%d, enabled=%d",
i, at.mediaType.rawValue as NSString, at.trackID, track.isEnabled)
} else {
NSLog("[Recorder] track[%d]: assetTrack=NIL (HLS not loaded yet?)", i)
}
}
//
if let audioTrack = self.findAudioTrackID(in: playerItem) {
NSLog("[Recorder] ✓ Audio track immediately available: trackID=%d", audioTrack)
self.installTap(on: playerItem, trackID: audioTrack, audioWriterInput: audioWriterInput)
return
}
// tracks KVO
NSLog("[Recorder] ⏳ Audio tracks not yet loaded, setting up KVO + polling...")
let observation = playerItem.observe(\.tracks, options: [.new]) { [weak self] item, _ in
guard let self, self.isRunning, self.audioTap == nil else { return }
let tCount = item.tracks.count
NSLog("[Recorder] KVO tracks changed: count=%d", tCount)
if let trackID = self.findAudioTrackID(in: item) {
NSLog("[Recorder] ✓ Audio track via KVO: trackID=%d", trackID)
Task { @MainActor in
guard self.isRunning, self.audioTap == nil else { return }
self.installTap(on: item, trackID: trackID, audioWriterInput: audioWriterInput)
}
self.tracksObservation = nil
}
}
self.tracksObservation = observation
// async asset.load(.tracks)
// 20
Task {
var attempts = 0
var lastAsyncAttempt = -100 //
while attempts < 200 && self.isRunning && self.audioTap == nil { // 20
try await Task.sleep(nanoseconds: 100_000_000) // 0.1
attempts += 1
// 1: playerItem.tracks
if let trackID = self.findAudioTrackID(in: playerItem) {
NSLog("[Recorder] ✓ Audio track via polling (attempt %d): trackID=%d", attempts, trackID)
await MainActor.run {
guard self.isRunning, self.audioTap == nil else { return }
self.installTap(on: playerItem, trackID: trackID, audioWriterInput: audioWriterInput)
}
self.tracksObservation = nil
return
}
// 2: 3 async load asset tracksHLS
if attempts - lastAsyncAttempt >= 30 {
lastAsyncAttempt = attempts
NSLog("[Recorder] Trying async asset.load(.tracks) (attempt at %d)...", attempts)
do {
let assetTracks = try await playerItem.asset.load(.tracks)
NSLog("[Recorder] async load returned %d tracks", assetTracks.count)
for (i, t) in assetTracks.enumerated() {
NSLog("[Recorder] assetTrack[%d]: mediaType=%@, trackID=%d",
i, t.mediaType.rawValue as NSString, t.trackID)
}
// track
if let audioAssetTrack = assetTracks.first(where: { $0.mediaType == .audio }) {
let trackID = audioAssetTrack.trackID
NSLog("[Recorder] ✓ Audio track via async load: trackID=%d", trackID)
await MainActor.run {
guard self.isRunning, self.audioTap == nil else { return }
self.installTap(on: playerItem, trackID: trackID, audioWriterInput: audioWriterInput)
}
self.tracksObservation = nil
return
}
// trackmuxed
if let muxedTrack = assetTracks.first(where: { $0.mediaType != .video && $0.mediaType != .audio }) {
let trackID = muxedTrack.trackID
NSLog("[Recorder] ✓ Muxed track via async load: trackID=%d, trying as audio", trackID)
await MainActor.run {
guard self.isRunning, self.audioTap == nil else { return }
self.installTap(on: playerItem, trackID: trackID, audioWriterInput: audioWriterInput)
}
self.tracksObservation = nil
return
}
} catch {
NSLog("[Recorder] async load failed: %@", error.localizedDescription)
}
}
// 5
if attempts % 50 == 0 {
let tc = playerItem.tracks.count
NSLog("[Recorder] polling attempt %d: playerItem.tracks.count=%d, tapInstalled=%d",
attempts, tc, self.audioTap != nil ? 1 : 0)
}
}
if self.audioTap == nil {
NSLog("[Recorder] ❌ No audio track found after %d polling attempts (20s timeout)", attempts)
self.tracksObservation = nil
}
}
}
/// playerItem.tracks trackID
/// HLS fallbackplayerItem.tracks asset.tracks
private func findAudioTrackID(in playerItem: AVPlayerItem) -> CMPersistentTrackID? {
// 1: playerItem.tracks + assetTrack / HLS
let itemTracks = playerItem.tracks
for track in itemTracks {
if let assetTrack = track.assetTrack, assetTrack.mediaType == .audio {
NSLog("[Recorder] findTrack: audio via playerItem.tracks, trackID=%d", assetTrack.trackID)
return assetTrack.trackID
}
}
// 2: asset.tracksHLS
let assetTracks = playerItem.asset.tracks
for track in assetTracks {
if track.mediaType == .audio {
NSLog("[Recorder] findTrack: audio via asset.tracks, trackID=%d", track.trackID)
return track.trackID
}
}
// 3: track track muxed
for track in itemTracks {
if let assetTrack = track.assetTrack,
assetTrack.mediaType != .video, assetTrack.mediaType != .audio {
NSLog("[Recorder] findTrack: non-video track (mediaType=%@, trackID=%d), trying as audio",
assetTrack.mediaType.rawValue as NSString, assetTrack.trackID)
return assetTrack.trackID
}
}
return nil
}
private func installTap(on playerItem: AVPlayerItem, trackID: CMPersistentTrackID, audioWriterInput: AVAssetWriterInput) {
//
let context = AudioCaptureContext()
context.writerInput = audioWriterInput
context.isRunning = true
self.audioContext = context
// C 访
gAudioContext = context
// MTAudioProcessingTap
var callbacks = MTAudioProcessingTapCallbacks(
version: kMTAudioProcessingTapCallbacksVersion_0,
clientInfo: nil,
init: tapInitCallback,
finalize: tapFinalizeCallback,
prepare: tapPrepareCallback,
unprepare: tapUnprepareCallback,
process: tapProcessCallback
)
var tap: MTAudioProcessingTap?
let status = MTAudioProcessingTapCreate(
kCFAllocatorDefault,
&callbacks,
kMTAudioProcessingTapCreationFlag_PostEffects,
&tap
)
guard status == noErr, let audioTap = tap else {
NSLog("[Recorder] ❌ MTAudioProcessingTapCreate failed: %d", status)
gAudioContext = nil
return
}
self.audioTap = audioTap
// playerItem trackID audioMix input parameters
let params = AVMutableAudioMixInputParameters()
params.trackID = trackID
params.audioTapProcessor = audioTap
let audioMix = AVMutableAudioMix()
audioMix.inputParameters = [params]
playerItem.audioMix = audioMix
NSLog("[Recorder] ✓ Audio tap installed on playerItem (trackID=%d)", trackID)
}
// MARK: - (30fps Timer)
private func startCaptureLoop() {
captureTimer = Timer.scheduledTimer(withTimeInterval: 1.0 / 30.0, repeats: true) { [weak self] _ in
Task { @MainActor in
self?.captureVideoFrame()
}
}
}
private func captureVideoFrame() {
guard isRunning, let output = weakOutput, let input = videoInput,
input.isReadyForMoreMediaData else { return }
let hostTime = CACurrentMediaTime()
let itemTime = output.itemTime(forHostTime: hostTime)
//
if let pb = output.copyPixelBuffer(forItemTime: itemTime, itemTimeForDisplay: nil) {
lastPixelBuffer = pb
}
// 使
guard let pb = lastPixelBuffer else { return }
//
let relativeTime = CMTimeSubtract(itemTime, recordStartTime)
guard relativeTime.seconds >= 0 else { return }
// sample buffer
if let sb = Self.createSampleBuffer(from: pb, time: relativeTime) {
if input.append(sb) {
captureFrameCount += 1
}
}
}
private static func createSampleBuffer(from pixelBuffer: CVPixelBuffer, time: CMTime) -> CMSampleBuffer? {
var info = CMSampleTimingInfo()
info.presentationTimeStamp = time
info.duration = CMTime(value: 1, timescale: 30)
info.decodeTimeStamp = .invalid
var formatDescription: CMFormatDescription?
CMVideoFormatDescriptionCreateForImageBuffer(
allocator: kCFAllocatorDefault,
imageBuffer: pixelBuffer,
formatDescriptionOut: &formatDescription
)
guard let fd = formatDescription else { return nil }
var sampleBuffer: CMSampleBuffer?
CMSampleBufferCreateReadyWithImageBuffer(
allocator: kCFAllocatorDefault,
imageBuffer: pixelBuffer,
formatDescription: fd,
sampleTiming: &info,
sampleBufferOut: &sampleBuffer
)
return sampleBuffer
}
// MARK: -
private func startDurationTimer() {
startDate = Date()
durationText = "00:00"
durationTimer = Timer.scheduledTimer(withTimeInterval: 1.0, repeats: true) { [weak self] _ in
Task { @MainActor in
guard let self = self, let start = self.startDate else { return }
let elapsed = Int(Date().timeIntervalSince(start))
let m = elapsed / 60
let s = elapsed % 60
let h = m / 60
if h > 0 {
self.durationText = String(format: "%d:%02d:%02d", h, m % 60, s)
} else {
self.durationText = String(format: "%02d:%02d", m, s)
}
}
}
}
private func stopDurationTimer() {
durationTimer?.invalidate()
durationTimer = nil
startDate = nil
durationText = "00:00"
}
// MARK: -
private func generateFinalFilename() -> String {
let formatter = DateFormatter()
formatter.dateFormat = "yyyy-MM-dd_HH-mm-ss"
return "Recording_\(formatter.string(from: Date())).mp4"
}
// MARK: -
private func showSaveDialog(tempURL: URL) {
let panel = NSSavePanel()
panel.nameFieldStringValue = generateFinalFilename()
panel.allowedContentTypes = [.mpeg4Movie]
panel.directoryURL = saveDirectory
panel.title = "保存录制文件"
panel.prompt = "保存"
panel.begin { [weak self] response in
Task { @MainActor in
guard let self = self else {
if RecorderState.pendingTerminate {
RecorderState.pendingTerminate = false
NSApp.reply(toApplicationShouldTerminate: true)
}
return
}
if response == .OK, let finalURL = panel.url {
do {
//
if FileManager.default.fileExists(atPath: finalURL.path) {
try FileManager.default.removeItem(at: finalURL)
}
//
try FileManager.default.moveItem(at: tempURL, to: finalURL)
NSLog("[Recorder] ✓ Saved: %@", finalURL.path)
self.onRecordingSaved?(finalURL)
} catch {
self.onError?("Save failed: \(error.localizedDescription)")
try? FileManager.default.removeItem(at: tempURL)
}
} else {
//
NSLog("[Recorder] User cancelled save, deleting temp file")
try? FileManager.default.removeItem(at: tempURL)
}
self.checkPendingTerminate()
}
}
}
private func checkPendingTerminate() {
if RecorderState.pendingTerminate {
RecorderState.pendingTerminate = false
NSLog("[MiniPlayer] Pending terminate triggered, replying to terminate")
NSApp.reply(toApplicationShouldTerminate: true)
}
}
}
#endif