If new photo is added to library and app is not running in foreground or was not opened after the new photo was added but the app is having full access to gallery, can it access, read the new photo - If the app is not specifically a cloud syncing app, can it have this attached function, suppose it is a game app or beauty camera app?

Hi all, with my app ScreenFloat, you can record your screen, along with system- and microphone audio. Those two audio feeds are recorded into separate audio tracks in order to individually remove or edit them later on. Now, these recordings you create with ScreenFloat can be drag-and-dropped to other apps instantly. So far, so good, but some apps, like Slack, or VLC, or even websites like YouTube, do not play back multiple audio tracks, just one. So what I'm trying to do is, on dragging the video recording file out of ScreenFloat, instantly baking together the two individual audio tracks into one, and offering that new file as the drag and drop file, so that all audio is played in the target app. But it's slow. I mean, it's actually quite fast, but for drag and drop, it's slow. My approach is this: "Bake together" the two audio tracks into a one-track m4a audio file using AVMutableAudioMix and AVAssetExportSession Take the video track, add the new audio file as an audio track to it, and render that out using AVAssetExportSession For a quick benchmark, a 3'40'' movie, step 1 takes ~1.7 seconds, and step two adds another ~1.5 seconds, so we're at ~3.2 seconds. That's an eternity for a drag and drop, where the user might cancel if there's no immediate feedback. I could also do it in one step, but then I couldn't use the AV*Passthrough preset, and that makes it take around 32 seconds then, because I assume it touches the video data (which is unnecessary in this case, so I think the two-step approach here is the fastest). So, my question is, is there a faster way? The best idea I can come up with right now is, when initially recording the screen with system- and microphone audio as separate tracks, to also record both of them into a third, muted, "hidden" track I could use later on, basically eliminating the need for step one and just ripping the two single audio tracks out of the movie and only have the video and the "hidden" track (then unmuted), but I'd still have a ~1.5 second delay there. Also, there's the processing and data overhead (basically doubling the movie's audio data). All this would be great for an export operation (where one expects it to take a little time), but for a drag-and-drop operation, it's not ideal. I've discarded the idea of doing a promise file drag, because many apps do not accept those, and I want to keep wide compatibility with all sorts of apps. I'd appreciate any ideas or pointers. Thank you kindly, Matthias

Hi, when using ApplicationMusicPlayer from MusicKit my app automatically gets the media controls on the lock screen: Play/ Pause, Skip Buttons, Playback Position etc. I would like to customize these. Tried a bunch of things, e.g. using MPRemoteCommandCenter. So far I haven't had any success. Does anyone know how I can customize the media controls of ApplicationMusicPlayer. Thank you.

Hello everyone, I'm implementing the new AVInputPickerInteraction API on iOS 26 to allow users to select their microphone from a custom settings menu before recording. The implementation seems correct, but I'm encountering a strange issue where the input selection immediately reverts to the previous device. The Situation: The picker is presented correctly via a manual call to .present(). I can see all available inputs (e.g., "iPhone Microphone" and "AirPods"). The current input is "iPhone Microphone". I tap on "AirPods". The UI updates to show "AirPods" as selected for a fraction of a second, then immediately jumps back to "iPhone Microphone". The same thing happens in reverse. It seems like the system is automatically reverting the audio route change requested by the picker. My Implementation: My setup follows the standard pattern discussed in the WWDC sessions. Setup Code: This setup is performed once before the user can trigger the picker. @available(iOS 26.0, *) var inputPickerInteraction: AVInputPickerInteraction? // Note: The AVAudioSession is configured to .playAndRecord // and set to active elsewhere in the code before this setup is called. if #available(iOS 26.0, *) { // Setup the picker let picker = AVInputPickerInteraction() self.inputPickerInteraction = picker self.view.addInteraction(picker) // Added to establish context } Presentation Code: When a user selects "Change Input" from my custom settings menu, I call .present() on the main thread. // In a delegate method from a custom menu if #available(iOS 26.0, *) { DispatchQueue.main.async { self.inputPickerInteraction?.present(animated: true) } } What I've already checked: The AVAudioSession is active and its category is .playAndRecord. The inputPickerInteraction object is not nil. The .present() method is being called on the main thread. The picker is added to a view using view.addInteraction() in the setup phase. I've reviewed my code to ensure there is no other logic that could be manually resetting the AVAudioSession's preferred input. Has anyone else experienced this behavior? I suspect this might be a bug in the new API, but I want to make sure I'm not missing a crucial step in managing the AVAudioSession state. Any insights or potential workarounds would be greatly appreciated. Thank you.

Hi everyone, I’m running into an issue with PHPickerFilter when using PHPickerViewController. When I configure the picker with a .videos and .livePhotos filter, it seems to work correctly in the Photos tab. However, when I switch to the Collections tab, the filter doesn’t always apply — users can still see and select static image assets in certain collections (e.g. from one of the People & Pets sections). Here’s a simplified snippet of my setup: var configuration = PHPickerConfiguration(photoLibrary: .shared()) configuration.selectionLimit = 1 var filters = [PHPickerFilter]() filters.append(.videos) filters.append(.livePhotos) configuration.filter = PHPickerFilter.any(of: filters) configuration.preferredAssetRepresentationMode = .current let picker = PHPickerViewController(configuration: configuration) picker.delegate = self present(picker, animated: true) Expected behavior: The picker should consistently respect the filter across both Photos and Collections tabs, only showing assets that match the filter. Actual behavior: The filter seems to apply correctly in the Photos tab, but in the Collections tab, other asset types are still visible/selectable. Has anyone else encountered this behavior? Is this expected or a known issue, or am I missing something in the configuration? Thanks in advance!

Hello, I have an iOS app that is recording audio that is working fine on iPads/iPhones. It asks for microphone permission and after that recording works. I installed the same app on my M3 MacBook via TestFlight since iPad apps are supposed to work without a change that way. The app starts fine and everything, but it never asks for Microphone permission, so I can't record. Do I need to do something to make this happen (this is not macCatalyst, its running the arm64 iPhone binary on macOS) thanks

The presentation "create audio drivers with DriverKit" from WWDC 2021 demonstrates how to use a dext to implement a virtual audio driver. It also says " If a virtual audio driver or device is all that is needed, the audio server plug-in driver model should continue to be used". Indeed, in AudioDriverKit/AudioDriverKitTypes.h, there is no IOUserAudioTransportType Virtual, although CoreAudio/AudioHardwareBase.h includes kAudioDeviceTransportTypeVirtual. For one of our products, we require virtual devices to implement a software loopback "cable". We've implemented this using the "traditional" HAL plugin, and as a proof-of-concept, also using a dext. In the dext, I tried setting the transport type to 'virt', which seems to only have the effect of changing the icon shown in Audio Midi Setup. HAL plugins require an installer, and the installer has to kill coreaudiod in a post-install script. You have to turn off SIP to debug them. Just like AudioDriverKit drivers, they are out-of-process and run in a process not owned by the hosting app. Our HAL plugin's interface is property based; we had to write a lot of boiler-plate code to implement required properties. Writing an AudioDriverKit driver is in most respects easier - a lot of the scaffolding is implemented in the base driver, which we only alter where required. Debugging and installation is much easier. The dext works just fine, as far as we can ascertain, just as well as a HAL plugin. So, my question is - is the advice to use a HAL plugin for a virtual device still correct in 2025? And if so, what's the objection? We'd really prefer to ship the AudioDriverKit virtual audio device.

Issue: After iOS 18.5 release, our app is experiencing a significant increase in AVCaptureSessionInterruptionReason.videoDeviceNotAvailableWithMultipleForegroundApps errors. Details: Our camera-related code has not been updated recently.However, we've observed that the error rate has significantly increased starting from May 2025. The error rate has risen from approximately 0.02% (2 in 10,000 users) to 0.1% (1 in 1,000 users). This represents a 5x increase in error occurrence. The frequency has increased noticeably since iOS 18.5 This is affecting our app's camera functionality and user experience Questions: Are there any known changes in iOS 18.5 regarding camera access management? What are the recommended best practices to handle this interruption reason? Are there any API changes we should be aware of? Best, Shay

I have an app under development - demo here - https://youtu.be/VbAfUk_eYl0?si=s6EDBx-4G6P_QbZO - which is sort of an audio player for airdropped files - something useful to musicians who dump work in progress to their phone, make notes, revise and update. I've been testing my handling of audio session interruption notifications, but seems to be a lot of inconsistency in how, when and why iOS delivers them, and I'm wondering if there is some rhyme or reason to it that I'm just not detecting. For example, I am playing a song in my app. Switch to Apple Music and start playing a song there. My app gets an interruption began notification - this is consistent. Switch back to my app, and about half the time, I will get an interruption ended notification (coupled often with a blast of the tail of whatever audio buffer was partially played when the interruption started, even though the engine was stopped - and followed by call to my AVAudioPlayerNodeCompletionCallback - is there some way to avoid this?). Half the time I don't get an interruption ended notification; my app can (as expected) end the interruption by activating the AVAudioSession and playing something. I have not been able to determine any pattern to this behavior, other than that if my app started playing using AVAudioPlayerNode.scheduleSegment rather than scheduleFile I think the notification will be consistently delivered on app activation rather than when I activate the session programmatically. I would like my app to behave deterministically, and would appreciate any help in deciphering what causes the inconsistent behavior in notifications from iOS.

I developed a driverkit extension based on overriding-the-default-usb-video-class-extension, but the link didn’t give the details of realization. I asked DTS who gave two tips: 1, Do you also have a CMIO extension to load in place of the default overriding-the-default-usb-video-class-extension 2, Your DriverKit extension’s info.plist is also missing the CameraAssistantBundleID. I want to know why a driverkit extension needs a CMIO extension, what’s the data and control flow?

I'm working on a photo app and I want to allow the user to display, edit and delete photos. I can fetch all photos using PHAsset.fetchAssets(with: options). This works as intended. However, I can't seem to find a way to prevent the user from seeing photos from a Shared Library. The PHAssetSourceType only contains typeCloudShared to only show items from a specific album; not library. How can I filter by iCloud Shared Library?

Hi, our CourAudio server plugin utilizes the SystemConfiguration.framework to store and restore specific shared system wide settings. While our application can authenticate to utilize the SystemConfiguration.framework to gain write access to the shared configuration settings the CoreAudio server plugin obviously can't have any user interaction and therefor does not authenticate. Is it possible to authenticate the CoreAudio server plugin to gain write permissions? Are there any entitlements or other means that would allow this? Thanks!

Is there a recommended way on macOS 26 Tahoe to take a CoreAudio AudioObjectID and use it to lookup the underlying USB LocationID? I previously used AudioObjectID to query the corresponding DeviceUID with kAudioDevicePropertyDeviceUID. Then I queried for the IOService matching kIOAudioEngineClassName with property kIOAudioEngineGlobalUniqueIDKey matching DeviceUID, and I loaded kUSBDevicePropertyLocationID from the result. This fails on macOS 26, because the IO Registry for the device has an entry for usbaudiod rather than AppleUSBAudioEngine, and usbaudiod does not include a kIOAudioEngineGlobalUniqueIDKey property (or any other property to map it to a CoreAudio DeviceUID). My use-case here is a piece of audio recording software that allows configuring a set of supported audio devices via USB HID prior to recording. I present the user with a list of CoreAudio devices to use, but without a way to lookup the underlying USB LocationID, I cannot guarantee that the configured device matches the selected device (e.g. if the user plugged in two identical microphones).

In iOS 26 (Developer Beta), the AVCaptureMetadataOutputObjectsDelegate no longer receives callbacks when metadataOutput.metadataObjectTypes = [.face] is set. On earlier iOS versions the issue does not occur. Interestingly, face detection works if I set the sessionPreset to .medium, but not with .high — except on the iPhone 16 Pro Max, where it works regardless.

Hello, Environment macOS 15.6.1 / Xcode 26 beta 7 / iOS 26 Beta 9 In a simple AVFoundation video-playback sample, I’m seeing different behavior between iOS 18 and iOS 26 regarding AVPlayerItem.didPlayToEndTimeNotification. I’ve attached a minimal sample below. Please replace videoURL with a valid short video URL. Repro steps Tap “Play” to start playback and let the video finish. The AVPlayerItem.didPlayToEndTimeNotification registered with NotificationCenter should fire, and you should see Play finished. in the console. Without relaunching, tap “Play” again. This is where the issue arises. Observed behavior On iOS 18 and earlier: The video does not play again (it does not restart from the beginning), but AVPlayerItem.didPlayToEndTimeNotification is posted and Play finished. appears in the console. The same happens every time you press “Play”. On iOS 26: Pressing “Play” does not post AVPlayerItem.didPlayToEndTimeNotification. The code path that prints Play finished. is never called (the callback enclosing that line is not invoked again). Building the same program with Xcode 16.4 and running it on an iOS 26 beta device shows the same phenomenon, which suggests there has been a behavioral change for AVPlayerItem.didPlayToEndTimeNotification on iOS 26. I couldn’t find any mention of this in the release notes or API Reference. Because the semantics around AVPlayerItem.didPlayToEndTimeNotification appear to differ, we’re forced to adjust our logic. If there is a way to achieve the iOS 18–style behavior on iOS 26, I would appreciate guidance. Alternatively, if this change is intentional, could you share the reasoning? Is iOS 26 the correct behavior from Apple’s perspective and iOS 18 (and earlier) behavior considered incorrect? Any official clarification would be extremely helpful. import UIKit import AVFoundation final class ViewController: UIViewController { private let videoURL = URL(string: "https://......mp4")! private var player: AVPlayer? private var playerItem: AVPlayerItem? private var playerLayer: AVPlayerLayer? private var observeForComplete: NSObjectProtocol? // UI private let playerContainerView = UIView() private let playButton = UIButton(type: .system) private let stopButton = UIButton(type: .system) private let replayButton = UIButton(type: .system) deinit { if let observeForComplete { NotificationCenter.default.removeObserver(observeForComplete) } } override func viewDidLoad() { super.viewDidLoad() view.backgroundColor = .systemBackground setupUI() setupPlayer() } override func viewDidLayoutSubviews() { super.viewDidLayoutSubviews() playerLayer?.frame = playerContainerView.bounds } // MARK: - Setup private func setupUI() { playerContainerView.translatesAutoresizingMaskIntoConstraints = false playerContainerView.backgroundColor = .black view.addSubview(playerContainerView) // Buttons playButton.setTitle("Play", for: .normal) stopButton.setTitle("Pause", for: .normal) replayButton.setTitle("RePlay", for: .normal) [playButton, stopButton, replayButton].forEach { $0.titleLabel?.font = .systemFont(ofSize: 16, weight: .semibold) $0.translatesAutoresizingMaskIntoConstraints = false $0.contentEdgeInsets = UIEdgeInsets(top: 10, left: 16, bottom: 10, right: 16) } let stack = UIStackView(arrangedSubviews: [playButton, stopButton, replayButton]) stack.axis = .horizontal stack.spacing = 16 stack.alignment = .center stack.distribution = .equalCentering stack.translatesAutoresizingMaskIntoConstraints = false view.addSubview(stack) NSLayoutConstraint.activate([ playerContainerView.topAnchor.constraint(equalTo: view.safeAreaLayoutGuide.topAnchor, constant: 20), playerContainerView.leadingAnchor.constraint(equalTo: view.leadingAnchor), playerContainerView.trailingAnchor.constraint(equalTo: view.trailingAnchor), playerContainerView.heightAnchor.constraint(equalToConstant: 200), stack.topAnchor.constraint(equalTo: playerContainerView.bottomAnchor, constant: 20), stack.centerXAnchor.constraint(equalTo: view.centerXAnchor) ]) // Action playButton.addTarget(self, action: #selector(didTapPlay), for: .touchUpInside) stopButton.addTarget(self, action: #selector(didTapStop), for: .touchUpInside) replayButton.addTarget(self, action: #selector(didTapReplayFromStart), for: .touchUpInside) } private func setupPlayer() { // AVURLAsset -> AVPlayerItem → AVPlayer let asset = AVURLAsset(url: videoURL) let item = AVPlayerItem(asset: asset) self.playerItem = item let player = AVPlayer(playerItem: item) player.automaticallyWaitsToMinimizeStalling = true self.player = player let layer = AVPlayerLayer(player: player) layer.videoGravity = .resizeAspect playerContainerView.layer.addSublayer(layer) layer.frame = playerContainerView.bounds self.playerLayer = layer // Notification if let observeForComplete { NotificationCenter.default.removeObserver(observeForComplete) } if let playerItem { observeForComplete = NotificationCenter.default.addObserver( forName: AVPlayerItem.didPlayToEndTimeNotification, object: playerItem, queue: .main ) { [weak self] _ in guard self != nil else { return } Task { @MainActor in print("Play finished.") } } } } // MARK: - Actions @objc private func didTapPlay() { player?.play() } @objc private func didTapStop() { player?.pause() } // RePlay @objc private func didTapReplayFromStart() { player?.seek(to: .zero, toleranceBefore: .zero, toleranceAfter: .zero) { [weak self] _ in self?.player?.play() } } } I would greatly appreciate an official response from Apple engineering on whether this is an intentional change, a regression, or an API contract clarification, and what the recommended approach is going forward. Thank you.

On macOS Sequoia, I'm having the hardest time getting this basic audio output to work correctly. I'm compiling in XCode using C99, and when I run this, I get audio for a split second, and then nothing, indefinitely. Any ideas what could be going wrong? Here's a minimum code example to demonstrate: #include <AudioToolbox/AudioToolbox.h> #include <stdint.h> #define RENDER_BUFFER_COUNT 2 #define RENDER_FRAMES_PER_BUFFER 128 // mono linear PCM audio data at 48kHz #define RENDER_SAMPLE_RATE 48000 #define RENDER_CHANNEL_COUNT 1 #define RENDER_BUFFER_BYTE_COUNT (RENDER_FRAMES_PER_BUFFER * RENDER_CHANNEL_COUNT * sizeof(f32)) void RenderAudioSaw(float* outBuffer, uint32_t frameCount, uint32_t channelCount) { static bool isInverted = false; float scalar = isInverted ? -1.f : 1.f; for (uint32_t frame = 0; frame < frameCount; ++frame) { for (uint32_t channel = 0; channel < channelCount; ++channel) { // series of ramps, alternating up and down. outBuffer[frame * channelCount + channel] = 0.1f * scalar * ((float)frame / frameCount); } } isInverted = !isInverted; } AudioStreamBasicDescription coreAudioDesc = { 0 }; AudioQueueRef coreAudioQueue = NULL; AudioQueueBufferRef coreAudioBuffers[RENDER_BUFFER_COUNT] = { NULL }; void coreAudioCallback(void* unused, AudioQueueRef queue, AudioQueueBufferRef buffer) { // 0's here indicate no fancy packet magic AudioQueueEnqueueBuffer(queue, buffer, 0, 0); } int main(void) { const UInt32 BytesPerSample = sizeof(float); coreAudioDesc.mSampleRate = RENDER_SAMPLE_RATE; coreAudioDesc.mFormatID = kAudioFormatLinearPCM; coreAudioDesc.mFormatFlags = kLinearPCMFormatFlagIsFloat | kLinearPCMFormatFlagIsPacked; coreAudioDesc.mBytesPerPacket = RENDER_CHANNEL_COUNT * BytesPerSample; coreAudioDesc.mFramesPerPacket = 1; coreAudioDesc.mBytesPerFrame = RENDER_CHANNEL_COUNT * BytesPerSample; coreAudioDesc.mChannelsPerFrame = RENDER_CHANNEL_COUNT; coreAudioDesc.mBitsPerChannel = BytesPerSample * 8; coreAudioQueue = NULL; OSStatus result; // most of the 0 and NULL params here are for compressed sound formats etc. result = AudioQueueNewOutput(&coreAudioDesc, &coreAudioCallback, NULL, 0, 0, 0, &coreAudioQueue); if (result != noErr) { assert(false == "AudioQueueNewOutput failed!"); abort(); } for (int i = 0; i < RENDER_BUFFER_COUNT; ++i) { uint32_t bufferSize = coreAudioDesc.mBytesPerFrame * RENDER_FRAMES_PER_BUFFER; result = AudioQueueAllocateBuffer(coreAudioQueue, bufferSize, &(coreAudioBuffers[i])); if (result != noErr) { assert(false == "AudioQueueAllocateBuffer failed!"); abort(); } } for (int i = 0; i < RENDER_BUFFER_COUNT; ++i) { RenderAudioSaw(coreAudioBuffers[i]->mAudioData, RENDER_FRAMES_PER_BUFFER, RENDER_CHANNEL_COUNT); coreAudioBuffers[i]->mAudioDataByteSize = coreAudioBuffers[i]->mAudioDataBytesCapacity; AudioQueueEnqueueBuffer(coreAudioQueue, coreAudioBuffers[i], 0, 0); } AudioQueueStart(coreAudioQueue, NULL); sleep(10); // some time to hear the audio AudioQueueStop(coreAudioQueue, true); AudioQueueDispose(coreAudioQueue, true); return 0; }

I'm able to get text to speech to audio file using the following code for iOS 12 iPhone 8 to create a car file: audioFile = try AVAudioFile( forWriting: saveToURL, settings: pcmBuffer.format.settings, commonFormat: .pcmFormatInt16, interleaved: false) where pcmBuffer.format.settings is: [AVAudioFileTypeKey: kAudioFileMP3Type, AVSampleRateKey: 48000, AVEncoderBitRateKey: 128000, AVNumberOfChannelsKey: 2, AVFormatIDKey: kAudioFormatLinearPCM] However, this code does not work when I run the app in iOS 18 on iPhone 13 Pro Max. The audio file is created, but it doesn't sound right. It has a lot of static and it seems the speech is very low pitch. Can anyone give me a hint or an answer?

Has Objective-C been deprecated?

Since iOS/iPadOs/tvOS 18 then we have run into a new problem with streaming of FairPlay encrypted video. On the affected streams then the audio plays perfectly but the video freezes for periods of a few seconds, so it will freeze for 5s or so, then be OK for a few seconds then freeze again. It is entirely reproducible when all the following are true the video streams were produced by a particular encoder (or particular settings, not sure on that) the video must be encrypted device is running some variety of iOS 18 (or iPadOS or tvOS) the device is an affected device Known devices are AppleTV 4K 2nd Gen iPad Pro 11" 1st and 2nd gen Devices known not to show the problem are all other AppleTV models iPhone 13 Pro and 16 Pro If we stream the same content, but unencrypted, then it plays perfectly, or if you play the encrypted stream on, say, tvOS 17. When the freezing occurs then we can see in the console logs repeating blocks of lines like the following default 18:08:46.578582+0000 videocodecd AppleAVD: AppleAVDDecodeFrameResponse(): Frame# 5771 DecodeFrame failed with error 0x0000013c default 18:08:46.578756+0000 videocodecd AppleAVD: AppleAVDDecodeFrameInternal(): failed - error: 316 default 18:08:46.579018+0000 videocodecd AppleAVD: AppleAVDDecodeFrameInternal(): avdDec - Frame# 5771, DecodeFrame failed with error: 0x13c default 18:08:46.579169+0000 videocodecd AppleAVD: AppleAVDDisplayCallback(): Asking fig to drop frame # 5771 with err -12909 - internalStatus: 315 also more relevant looking lines: default 18:17:39.122019+0000 kernel AppleAVD: avdOutbox0ISR(): FRM DONE (cid: 2.0, fno: 10970, codecT: 1) FAILED!! default 18:17:39.122155+0000 videocodecd AppleAVD: AppleAVDDisplayCallback(): Asking fig to drop frame # 10970 with err -12909 - internalStatus: 315 default 18:17:39.122221+0000 kernel AppleAVD: ## client[ 2.0] @ frm 10970, errStatus: 0x10 default 18:17:39.122338+0000 kernel AppleAVD: decodeFailIdentify(): VP error bit 4 has EP3B0 error default 18:17:39.122401+0000 kernel AppleAVD: processHWResponse(): clientID 2.0 frameNumber 10970 error 315, offsetIndex 10, isHwErr 1 So it would seem to me that one of the following must be happening: When these particular HLS files are encrypted then the data is being corrupted in some way that played back on iOS 17 and earlier but now won't on 18+, or There's a regression in iOS 18 that means that this particular format of video data is corrupted on decryption If anyone has seen similar behaviour, or has any ideas how to identify which of the two scenarios it is, please say. Unfortunately we don't have control of the servers so can't make changes there unless we can identify they are definitely the cause of the problem. Thanks, Simon.

I'm writing some camera functionality that uses AVCaptureVideoDataOutput. I've set it up so that it calls my AVCaptureVideoDataOutputSampleBufferDelegate on a background thread, by making my own dispatch_queue and configuring the AVCaptureVideoDataOutput. My question is then, if I configure my AVCaptureSession differently, or even stop it altogether, is this guaranteed to flush all pending jobs on my background thread? For example, does [AVCaptureSession stopRunning] imply a blocking call until all pending frame-callbacks are done? I have a more practical example below, showing how I am accessing something from the foreground thread from the background thread, but I wonder when/how it's safe to clean up that resource. I have setup similar to the following: // Foreground thread logic dispatch_queue_t queue = dispatch_queue_create("qt_avf_camera_queue", nullptr); AVCaptureSession *captureSession = [[AVCaptureSession alloc] init]; setupInputDevice(captureSession); // Connects the AVCaptureDevice... // Store some arbitrary data to be attached to the frame, stored on the foreground thread FrameMetaData frameMetaData = ...; MySampleBufferDelegate *sampleBufferDelegate = [MySampleBufferDelegate alloc]; // Capture frameMetaData by reference in lambda [sampleBufferDelegate setFrameMetaDataGetter: [&frameMetaData]() { return &frameMetaData; }]; AVCaptureVideoDataOutput *captureVideoDataOutput = [[AVCaptureVideoDataOutput alloc] init]; [captureVideoDataOutput setSampleBufferDelegate:sampleBufferDelegate queue:queue]; [captureSession addOutput:captureVideoDataOutput]; [captureSession startRunning]; [captureSession stopRunning]; // Is it now safe to destroy frameMetaData, or do we need manual barrier? And then in MySampleBufferDelegate: - (void)captureOutput:(AVCaptureOutput *)captureOutput didOutputSampleBuffer:(CMSampleBufferRef)sampleBuffer fromConnection:(AVCaptureConnection *)connection { // Invokes the callback set above FrameMetaData *frameMetaData = frameMetaDataGetter(); emitSampleBuffer(sampleBuffer, frameMetaData); }