MiniPlayer/Sources/PlayerRecorder.swift
yumoqing be0cf09941 fix: HLS audio (playerItem.tracks + better tap callback) + window title sync with toolbar
- HLS录制音频: 改用playerItem.tracks查找trackID(fallback到asset.loadTracks)
- tap回调: 始终先调GetSourceAudio保证管道畅通,加详细日志
- 窗口标题: onAppear初始隐藏,随toolbar一起30秒后消失/点击图标回来
2026-06-25 09:08:02 +08:00

571 lines
21 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
}
// 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
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", asbd.mSampleRate, asbd.mChannelsPerFrame)
}
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 = Double(timeRange.duration.timescale) > 0
? Double(timeRange.duration.timescale) : 44100.0
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
/// 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?
//
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
captureTimer?.invalidate()
captureTimer = nil
stopDurationTimer()
// tap
audioContext?.isRunning = false
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
Task { @MainActor in
guard let self = self, let w = self.writer else { 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 {
//
self.showSaveDialog(tempURL: tempURL)
} else {
self.onError?("Recording failed: \(w.error?.localizedDescription ?? "unknown (status=\(w.status.rawValue))")")
if let url = self.tempURL {
try? FileManager.default.removeItem(at: url)
}
}
}
}
lastPixelBuffer = nil
}
// MARK: - Tap
private func setupAudioTap(playerItem: AVPlayerItem, audioWriterInput: AVAssetWriterInput) {
// track
// 1) playerItem.tracks tracktrackID audioMix
// 2) asset.loadTracksfallbackHLS tracks
Task {
var attempts = 0
// 1 playerItem.tracks 3
while attempts < 30 {
guard self.isRunning else { return }
let itemTracks = playerItem.tracks
if let audioItemTrack = itemTracks.first(where: { $0.assetTrack?.mediaType == .audio }),
let assetTrack = audioItemTrack.assetTrack {
let trackID = assetTrack.trackID
NSLog("[Recorder] ✓ Audio track via playerItem.tracks: trackID=%d", trackID)
self.installTap(on: playerItem, trackID: trackID, audioWriterInput: audioWriterInput)
return
}
attempts += 1
try await Task.sleep(nanoseconds: 100_000_000) // 0.1
}
// 2fallback asset.loadTracks 5
NSLog("[Recorder] ⚠️ playerItem.tracks empty, falling back to asset.loadTracks")
attempts = 0
while attempts < 50 {
guard self.isRunning else { return }
if let assetTracks = try? await playerItem.asset.loadTracks(withMediaType: .audio),
let audioTrack = assetTracks.first {
let trackID = audioTrack.trackID
NSLog("[Recorder] ✓ Audio track via asset.loadTracks: trackID=%d", trackID)
self.installTap(on: playerItem, trackID: trackID, audioWriterInput: audioWriterInput)
return
}
attempts += 1
try await Task.sleep(nanoseconds: 100_000_000) // 0.1
}
NSLog("[Recorder] ❌ No audio track found after both attempts")
}
}
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 { 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)
}
}
}
}
}
#endif