Hi, I'm developing an application for macos and ios that has to run DetectHumanBodyPose3DRequest model in real time for retrieving the 3d skeleton from the camera. I'm experiencing a memory leak every time the model is used (when i comment that line, the memory stays constant). After a minute it uses about 1GB of ram running with mac catalyst. I attached a minimal project that has this problem Code Camera View import SwiftUI import Combine import Vision struct CameraView: View { @StateObject private var viewModel = CameraViewModel() var body: some View { HStack { ZStack { GeometryReader { geometry in if let image = viewModel.currentFrame { Image(decorative: image, scale: 1) .resizable() .scaledToFill() .frame(width: geometry.size.width, height: geometry.size.height) .clipped() } else { ProgressView() } } } } } } class CameraViewModel: ObservableObject { @Published var currentFrame: CGImage? @Published var frameRate: Double = 0 @Published var currentVisionBodyPose: HumanBodyPose3DObservation? // Store current body pose @Published var currentImageSize: CGSize? // Store current image size private var cameraManager: CameraManager? private var humanBodyPose = HumanBodyPose3DDetector() private var lastClassificationTime = Date() private var frameCount = 0 private var lastFrameTime = Date() private let classificationThrottleInterval: TimeInterval = 1.0 private var lastPoseSendTime: Date = .distantPast init() { cameraManager = CameraManager() startPreview() startClassification() } private func startPreview() { Task { guard let previewStream = cameraManager?.previewStream else { return } for await frame in previewStream { let size = CGSize(width: frame.width, height: frame.height) Task { @MainActor in self.currentFrame = frame self.currentImageSize = size self.updateFrameRate() } } } } private func startClassification() { Task { guard let classificationStream = cameraManager?.classificationStream else { return } for await pixelBuffer in classificationStream { self.classifyFrame(pixelBuffer: pixelBuffer) } } } private func classifyFrame(pixelBuffer: CVPixelBuffer) { humanBodyPose.runHumanBodyPose3DRequestOnImage(pixelBuffer: pixelBuffer) { [weak self] observation in guard let self = self else { return } DispatchQueue.main.async { if let observation = observation { self.currentVisionBodyPose = observation print(observation) } else { self.currentVisionBodyPose = nil } } } } private func updateFrameRate() { frameCount += 1 let now = Date() let elapsed = now.timeIntervalSince(lastFrameTime) if elapsed >= 1.0 { frameRate = Double(frameCount) / elapsed frameCount = 0 lastFrameTime = now } } } HumanBodyPose3DDetector import Foundation import Vision class HumanBodyPose3DDetector: NSObject, ObservableObject { @Published var humanObservation: HumanBodyPose3DObservation? = nil private let queue = DispatchQueue(label: "humanbodypose.queue") private let request = DetectHumanBodyPose3DRequest() private struct SendablePixelBuffer: @unchecked Sendable { let buffer: CVPixelBuffer } public func runHumanBodyPose3DRequestOnImage(pixelBuffer: CVPixelBuffer, completion: @escaping (HumanBodyPose3DObservation?) -> Void) { let sendableBuffer = SendablePixelBuffer(buffer: pixelBuffer) queue.async { [weak self] in Task { [weak self, sendableBuffer] in do { guard let self = self else { return } let result = try await self.request.perform(on: sendableBuffer.buffer) //process result DispatchQueue.main.async { if result.isEmpty { completion(nil) } else { completion(result[0]) } } } catch { DispatchQueue.main.async { completion(nil) } } } } } }

Hi everyone, I’m trying to use AVAssetResourceLoaderDelegate to handle a live radio stream (e.g. Icecast/HTTP stream). My goal is to have access to the last 30 seconds of audio data during playback, so I can analyze it for specific audio patterns in near-real-time. I’ve implemented a custom resource loader that works fine for podcasts and static files, where the file size and content length are known. However, for infinite live streams, my current implementation stops receiving new loading requests after the first one is served. As a result, the playback either stalls or fails to continue. Has anyone successfully used AVAssetResourceLoaderDelegate with a continuous radio stream? Or maybe you can suggest betterapproach for buffering and analyzing live audio? Any tips, examples, or advice would be appreciated. Thanks!

Hi everyone, I've encountered a rare and strange crash in my app that I can't consistently reproduce. The crash seems to occur deep within Apple's internal frameworks, and I can't pinpoint which line of my own code is causing it. Here's the crash stack trace: #44 AXSpeech SIGSEGV SEGV_ACCERR 0 CoreFoundation ___CFCheckCFInfoPACSignature + 4 1 CoreFoundation _CFRunLoopSourceSignal + 28 2 Foundation _performQueueDequeue + 492 3 Foundation ___NSThreadPerformPerform + 88 4 CoreFoundation ___CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE0_PERFORM_FUNCTION__ + 28 5 CoreFoundation ___CFRunLoopDoSource0 + 176 6 CoreFoundation ___CFRunLoopDoSources0 + 340 7 CoreFoundation ___CFRunLoopRun + 828 8 CoreFoundation _CFRunLoopRunSpecific + 608 9 Foundation -[NSRunLoop(NSRunLoop) runMode:beforeDate:] + 212 10 TextToSpeech _TTSCFAttributedStringCreateStringByBracketingAttributeWithString + 776 11 Foundation ___NSThread__start__ + 732 12 libsystem_pthread.dylib __pthread_start + 136 Sometimes, instead of line 10 referencing _TTSCFAttributedStringCreateStringByBracketingAttributeWithString, it shows: 10 TextToSpeech LogWarning(char const*, ...) + 7288 Has anyone experienced a similar issue or know what might be triggering this crash? Any guidance on how to investigate or resolve this would be greatly appreciated. Thank you!

We use several UIKit and AVFoundation APIs in our project, including: setAlternateIconName(_:completionHandler:) getAllTasks(completionHandler:) loadMediaSelectionGroup(for:completionHandler:) Moreover, we use the Swift Concurrency versions for these APIs: @MainActor func setAlternateIconName(_ alternateIconName: String?) async throws var allTasks: [URLSessionTask] { get async } func loadMediaSelectionGroup(for mediaCharacteristic: AVMediaCharacteristic) async throws -> AVMediaSelectionGroup? Everything worked well with these APIs in Xcode 16.2 and earlier, but starting from Xcode 16.3 (and in 16.4), they cause crashes. We've rewritten the APIs to use completion blocks instead of async/await, and this approach works. Stack traces: setAlternateIconName(_:completionHandler:) var allTasks: [URLSessionTask] { get async } loadMediaSelectionGroup(for:completionHandler:) Also, I attached some screenshots from Xcode 16.4.

XCode 26 beta 2 is taking more than 20 seconds to start AVCaptureSession when AVCaptureVideoDataOutput and AVCaptureAudioDataOutput are added. The problem occurs only during debugging and is clearly seen with Cinematic Capture sample code by Apple. I am using iPhone 14 Pro running iOS 26 beta 2 for reference.

I have this build error with Xcode 26 beta 2: var asset:AVURLAsset? func loadAsset() { let assetURL = URL.documentsDirectory .appendingPathComponent("sample.mov") asset = AVURLAsset(url: assetURL, options: [AVURLAssetPreferPreciseDurationAndTimingKey: true]) /*Error: Type of expression is ambiguous without a type annotation */ if let result = try? await asset?.load(.tracks, .isPlayable, .isComposable) { } } Is there an issue with try? in the new Swift compiler? Error: Type of expression is ambiguous without a type annotation

We are facing a strange issue where a small portion of our large userbase can not start the capture session in our app, as it gets interrupted with the following reason: AVCaptureSessionInterruptionReasonVideoDeviceNotAvailableWithMultipleForegroundApps Our users are all from iPhones, no one is using an iPad. Just to be sure we have set session.isMultitaskingCameraAccessEnabled = true but it does not seem to make any difference. Another weird interruption we are seeing

We are facing a strange issue where a small portion of our large userbase can not start the capture session in our app, as it gets interrupted with the following reason: AVCaptureSessionInterruptionReasonVideoDeviceNotAvailableWithMultipleForegroundApps Our users are all from iPhones, no one is using an iPad. Just to be sure we have set session.isMultitaskingCameraAccessEnabled = true but it does not seem to make any difference. Another weird scenario we are seeing on an even smaller number of users is that the following call: AVCaptureDevice.default(.builtInWideAngleCamera, for: .video, position: .back) returns nil. A quick look at our error reports show this happening on iPhone XR, 13 and 14 models. They should all support this device type. Any help on investigating these issue would be greatly appreciated!

Hello, I'm trying to subscribe to AVPlayerItem status updates using Combine and it's bridge to Swift Concurrency – .values. This is my sample code. struct ContentView: View { @State var player: AVPlayer? @State var loaded = false var body: some View { VStack { if let player { Text("loading status: \(loaded)") Spacer() VideoPlayer(player: player) Button("Load") { Task { let item = AVPlayerItem( url: URL(string: "https://sample-videos.com/video321/mp4/360/big_buck_bunny_360p_5mb.mp4")! ) player.replaceCurrentItem(with: item) let publisher = player.publisher(for: \.status) for await status in publisher.values { print(status.rawValue) if status == .readyToPlay { loaded = true break } } print("we are out") } } } else { Text("No video selected") } } .task { player = AVPlayer() } } } After I click on the "load" button it prints out 0 (as the initial status of .unknown) and nothing after – even when the video is fully loaded. At the same time this works as expected (loading status is set to true): struct ContentView: View { @State var player: AVPlayer? @State var loaded = false @State var cancellable: AnyCancellable? var body: some View { VStack { if let player { Text("loading status: \(loaded)") Spacer() VideoPlayer(player: player) Button("Load") { Task { let item = AVPlayerItem( url: URL(string: "https://sample-videos.com/video321/mp4/360/big_buck_bunny_360p_5mb.mp4")! ) player.replaceCurrentItem(with: item) let stream = AsyncStream { continuation in cancellable = item.publisher(for: \.status) .sink { if $0 == .readyToPlay { continuation.yield($0) continuation.finish() } } } for await _ in stream { loaded = true cancellable?.cancel() cancellable = nil break } } } } else { Text("No video selected") } } .task { player = AVPlayer() } } } Is this a bug or something?

I'm seeking to a specific sync frame in a video file (HEVC, recorded on iPad). When I feed the buffers from that sync frame on to VTDecompressionSession it consistently drops the 2.,3.,4. buffer with a kVTVideoDecoderReferenceMissingErr (or no error but no buffer on the simulator). If I feed all the buffers from the penultimate sync frame prior to the desired frame the buffers come out fine but that would just create a massive overhead to always do it. Tried multiple OS versions, devices etc. Seems a consistent problem. Here's a sample project with the offending video (disregard memory handling etc): https://github.com/marcuseckert/vtSample I've filed a radar FB18228296 but would appreciate any feedback on circumventing or at least detecting this behavior prior to decoding.

We are seeing logs were on iOS devices we see some keyframes request. but on safari browser don’t see any request like this. could you please explain what is it. /d8ceb9244ff889b42b82eb807327531-c27dbcb10e0bbf3cde6c-1/d8ceb9244ff88e9b42b82eb807327531-c27dbcb10e0bbf3cde6c-1/keyframes/hls/.

TL;DR How to solve possible racing issue of EXT-X-SESSION-KEY request and encrypted media segment request? I'm having trouble using custom AVAssetResourceLoaderDelegate with video manifest containing VideoProtectionKey(VPK). My master manifest contains rendition manifest url and VPK url. When not using custom resource delegate, everything works fine. My custom resource delegate is implemented in way where it first append prefix to scheme of the master manifest url before creating the asset. And during handling master manifest, it puts back original scheme, make the request, modify the scheme for rendition manifest url in the response content by appending the same prefix again, so that rendition manifest request also goes into custom resource loader delegate. Same goes for VPK request. The AES-128 key is stored in memory within custom resource loader delegate object. So far so good. The VPK is requested before segment request. But the problem comes where the media segment requests happen. The media segment request url from rendition manifest goes into custom resource loader as well and those are encrypted. I can see segment request finish first then the related VPK requests kick in after a few seconds. The previous VPK value is cached in memory so it is not network causing the delay but some mechanism that I'm not aware of causing this. So could anyone tell me what would be the proper way of handling this situation? The native library is handling it well so I just want to know how. Thanks in advance!

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 am wondering if new AVCam sample code was tested before release. It hangs on startup on iPhone 14 pro running iOS 26 beta with the following logs on console: <<<< FigAudioSession(AV) >>>> audioSessionAVAudioSession_CopyMXSessionProperty signalled err=-19224 (kFigAudioSessionError_UnsupportedOperation) (getMXSessionProperty unsupported) at FigAudioSession_AVAudioSession.m:606 <<<< FigAudioSession(AV) >>>> audioSessionAVAudioSession_CopyMXSessionProperty signalled err=-19224 (kFigAudioSessionError_UnsupportedOperation) (getMXSessionProperty unsupported) at FigAudioSession_AVAudioSession.m:606 App is being debugged, do not track this hang Hang detected: 8.04s (debugger attached, not reporting)```

I'm seeing this error while attempting to compile my VisionOS app under Xcode 26. My existing code looks like: let (naturalSize, formatDescriptions, mediaCharacteristics) = try? await videoTrack.load(.naturalSize, .formatDescriptions, .mediaCharacteristics) This is now giving a compiler error: Type of expression is ambiguous without a type annotation I don't see that anything that was changed or deprecated in the latest version. Also loading the properties individually seems to work fine i.e.: let naturalSize = try? await videoTrack.load(.naturalSize) let formatDescriptions = try? await videoTrack.load(.formatDescriptions) let mediaCharacteristics = try? await videoTrack.load(.mediaCharacteristics)

Among the millions of users of our online product, we have identified through data metrics that the silent audio data capture rate on iPadOS 18.4.1 or 18.5 has increased abnormally. However, we are unable to reproduce the issue. Has anyone encountered a similar issue? The parameters we used are as follows: AudioSession: category:AVAudioSessionCategoryPlayAndRecord mode:AVAudioSessionModeDefault option:77 preferredSampleRate:48000.000000 preferredIOBufferDuration:0.010000 AudioUnit format.mFormatID = kAudioFormatLinearPCM; format.mSampleRate = 48000.0; format.mChannelsPerFrame = 2; format.mBitsPerChannel = 16; format.mFramesPerPacket = 1; format.mBytesPerFrame = format.mChannelsPerFrame * 16 / 8; format.mBytesPerPacket = format.mBytesPerFrame * format.mFramesPerPacket; format.mFormatFlags = kAudioFormatFlagsNativeEndian | kLinearPCMFormatFlagIsPacked | kLinearPCMFormatFlagIsSignedInteger; component.componentType = kAudioUnitType_Output; component.componentSubType = kAudioUnitSubType_RemoteIO; component.componentManufacturer = kAudioUnitManufacturer_Apple; component.componentFlags = 0; component.componentFlagsMask = 0;

Hello there! Is there any list of voices that are always available on iOS/iPadOS devices? It seems that AVSpeechSynthesisVoice(identifier: "com.apple.voice.compact.en-US.Samantha") is always available on all devices. I thought that AVSpeechSynthesisVoice(identifier: "com.apple.ttsbundle.siri_Nicky_en-US_compact") and AVSpeechSynthesisVoice(identifier: "com.apple.ttsbundle.siri_Aaron_en-US_compact") were available by default on certain newer devices. Is this true? I also noticed that on the same iPad where I was using those 2 voices (Nicky and Aaron) - when I updated to the iPadOS 26 beta, those voices were no longer available. Any information you can share about which voices should be reliably available on which devices would be extremely helpful for our development. Thanks so much!

Please consider adding the ability to programatically download Premium and Enhanced voices. At the moment it is extremely inconvenient for our users, as they have to navigate to settings themselves to download voices. Our app relies heavily on SpeechSynthesis integration, and it would greatly benefit from this feature. FB16307193

When I play an HDR video in the iPhone Photos app, I can see the HDR effect obviously. But if this HDR video is played continuously for more than 30-40 minutes, the HDR effect will disappear and the brightness will be compressed to the SDR range. This issue will appear on any iPhone. Depending on the phone, it may be 20-30 minutes, or 30-40 minutes, or even a few minutes, such as iPhone 12 mini. Similarly, if I use AVPlayer to play and preview an HDR video, if it plays more than 30-40 minutes, the HDR effect will disappear and the screen brightness will dim. Also the currentEDRHeadroom will gradually decrease to 1 Note, test it with an HDR video longer than 1 hour, and if the video is short, please loop it. My question is how to avoid losing the HDR effect after 30-40 minutes when I use CAMetalLayer to render any HDR video.

Hello, We're seeing an intermittent issue when playing back FairPlay-protected HLS downloads while the device is offline. Assets are downloaded using AVAggregateAssetDownloadTask with FairPlay protection. After download, asset.assetCache.isPlayableOffline == true. On first playback attempt (offline), ~8% of downloads fail. Retrying playback always works. We recreate the asset and player on each attempt. During the playback setup, we try to load variants via: try await asset.load(.variants) This call sometimes fails with: Error Domain=NSURLErrorDomain Code=-1009 “The Internet connection appears to be offline.” UserInfo={NSUnderlyingError=0x105654a00 {Error Domain=NSURLErrorDomain Code=-1009 “The Internet connection appears to be offline.” UserInfo={NSDescription=The Internet connection appears to be offline.}}, NSErrorFailingURLStringKey=file:///private/var/mobile/Containers/Data/Application/2DDF9D7C-9197-46BE-8690-C23EE75C9E90/Library/com.apple.UserManagedAssets.XVvqfh/Baggage_9DD4E2D3F9C0E68F.movpkg/, NSErrorFailingURLKey=file:///private/var/mobile/Containers/Data/Application/2DDF9D7C-9197-46BE-8690-C23EE75C9E90/Library/com.apple.UserManagedAssets.XVvqfh/Baggage_9DD4E2D3F9C0E68F.movpkg/, NSURL=file:///private/var/mobile/Containers/Data/Application/2DDF9D7C-9197-46BE-8690-C23EE75C9E90/Library/com.apple.UserManagedAssets.XVvqfh/Baggage_9DD4E2D3F9C0E68F.movpkg/, AVErrorFailedDependenciesKey=( “assetProperty_HLSAlternates” ), NSLocalizedDescription=The Internet connection appears to be offline.} This variant load is used to determine available audio tracks, check for Dolby support, and apply user language preferences. After this step, the AVPlayerItem also fails via Combine’s publisher for .status. However, retrying the entire process immediately after (same offline conditions, same asset path, new AVURLAsset) results in successful playback. Assets are represented using the following class: public class DownloadedAsset: AVURLAsset { public let id: String public let localFileUrl: URL public let fairplayLicenseUrlString: String? public let drmToken: String? var isProtected: Bool { return fairplayLicenseUrlString != nil } public init(id: String, localFileUrl: URL, fairplayLicenseUrlString: String?, drmToken: String?) { self.id = id self.localFileUrl = localFileUrl self.fairplayLicenseUrlString = fairplayLicenseUrlString self.drmToken = drmToken super.init(url: localFileUrl, options: nil) } } We use user-selected quality levels to control bitrate and multichannel (e.g. Dolby 5.1) downloads: let downloadQuality = UserDefaults.standard.downloadVideoQuality let bitrate: Int let shouldDownloadMultichannelTracks: Bool switch downloadQuality { case .dataSaver: shouldDownloadMultichannelTracks = false bitrate = 596564 case .standard: shouldDownloadMultichannelTracks = false bitrate = 1503844 case .best: shouldDownloadMultichannelTracks = true bitrate = 7038970 } var selections = multichannelIdentifiedMediaSelections if !shouldDownloadMultichannelTracks { selections = selections.filter { !$0.isMultichannel } } let task = session.aggregateAssetDownloadTask( with: asset, mediaSelections: selections.map { $0.mediaSelection }, assetTitle: title, assetArtworkData: nil, options: [AVAssetDownloadTaskMinimumRequiredMediaBitrateKey: bitrate] ) Seen on devices running iOS 16, iOS 17, and iOS 18. What could cause the initial failure of an otherwise valid, offline-ready FairPlay HLS asset? Could .load(.variants) internally trigger a failed network resolution, even when offline? Is there an internal caching or initialization behavior in AVFoundation that might explain why the second attempt works? Any guidance would be appreciated.