Dive into the technical aspects of audio on your device, including codecs, format support, and customization options.

Audio Documentation

Posts under Audio subtopic

Post

Replies

Boosts

Views

Activity

iPhone 18 Pro: AVAudioEngine input tap returns all zeros after the Always-On Display turns off overnight
My app records audio overnight in the background: UIBackgroundModes audio, AVAudioSession category playAndRecord with mode measurement, and an AVAudioEngine with a tap on inputNode. On a customer's iPhone 18 Pro running iOS 27.2.1 (24B156), the tap starts delivering buffers of exact zeros about two hours into the night. It keeps delivering zeros until she picks the phone up in the morning. Her sysdiagnose shows the same sequence on three consecutive nights: powerd logs SmartPowerNap: enterSmartPowerNap. The backlight state changes from inactiveOnLive to suppressed, with the explanation "alwaysOn suppressed - SmartPowerNap". The panel reaches 0 nits about two seconds later. Under a second after that, every sample in the recording is 0. PowerLog's per-app mic RMS reads -240 from that point on. Hours later powerd logs Exiting SmartPowerNap. Session end reason MotionDetected. The display wakes and mic input returns within a second. Timing on the third night: 23:55:57.234 Smart Power Nap entry 23:55:59.14 panel at 0 nits 23:55:59.8 first zero sample in the recording 05:46:58.032 Smart Power Nap exit about 05:46:59 mic input returns During the silent hours: The app receives no interruption, route change, media services reset or AVAudioEngineConfigurationChange notification. The engine keeps running and the tap keeps firing on schedule. audiomxd logs no deactivation for the session and holds its power assertions throughout. The system never sleeps. I can't reproduce it. My iPhone 16 Pro on the same build logs "SmartPowerNap: feature not enabled or predictor not initialized" on every lock, so it never takes this path. Questions: Does iOS gate mic input by design while Smart Power Nap suppresses the Always-On Display? Can an app keep recording through it, or get a notification that input has stopped? Does restarting AVAudioEngine while the session stays active restore input in this state? I can detect a run of exact zeros and restart the engine, but I have no device to test it on. Has anyone else with iPhone 18 Pro users seen this? I filed FB25663635 with the sysdiagnose attached.
0
0
73
1d
Playback volume stays low with PlayAndRecord/VoiceChat and remains low for a few seconds after restoring SoloAmbient (iPhone Air, iOS 26)
Hello, I'm developing a speech recognition app with Unity 6000.0.58f2. The app also has other content besides speech recognition, so I switch the AVAudioSession only when the user needs to speak, and restore it once recording is no longer needed. On iPhone Air (iOS 26), I'm seeing the following issues: While the session is set to PlayAndRecord with VoiceChat mode, playback volume is consistently lower than normal. After restoring the session to SoloAmbient, playback volume stays low for a few seconds before returning to normal. I'd like to avoid this volume drop. Any advice or recommended approach would be greatly appreciated. [Implementation] I use a native iOS plugin (.mm file) called from Unity. When the user starts speaking, I enable recording with: success = [session setCategory:AVAudioSessionCategoryPlayAndRecord mode:AVAudioSessionModeVoiceChat options:AVAudioSessionCategoryOptionDefaultToSpeaker | AVAudioSessionCategoryOptionAllowBluetooth error:&error];After speaking ends, I restore the session with: success = [session setCategory:AVAudioSessionCategorySoloAmbient error:&error]; success = [session setMode:AVAudioSessionModeDefault error:&error];[Environment] Device: iPhone Air iOS: 26.[x] Unity: 6000.0.58f2 Reproduces on other devices / iOS versions: [Yes / No — e.g. iPhone 15 (iOS 18.x) does not reproduce] Output route: built-in speaker [/ Bluetooth: ___ ] [Questions] Is the lower playback volume in VoiceChat mode expected behavior due to voice processing (echo cancellation / ducking)? If so, is there a recommended way to keep playback volume at a normal level while recording for speech recognition (e.g. using a different mode such as Default or Measurement, or adjusting voice processing ducking)? What is the recommended sequence for restoring the session so that playback volume returns to normal immediately? For example, should I call setActive:NO (with NotifyOthersOnDeactivation) before changing the category, or set the category and mode together in a single call? Is there any behavior change in iOS 26 or anything specific to iPhone Air that could cause this? Thank you in advance.
0
0
155
2d
Feature Request: Allow Independent Bluetooth Microphone Input and Speaker Output on iPhone
I hope iphone can consider adding this feature in a future update. We are now living in the era of content creation and social media. With the growing popularity of Bluetooth wireless microphones, more and more people are using external microphones for gaming, live streaming, video calls, and content creation. However, there is a frustrating limitation: when connecting a Bluetooth microphone, the phone may automatically route audio output to the same Bluetooth device. This means users cannot freely use an external microphone for input while keeping the phone's built-in speakers for audio output. What we really need is the ability to independently select audio input and output devices at the system level, just like on a PC. For example: Audio input: DJI wireless microphone Audio output: Phone's built-in speakers This would be extremely useful for mobile gamers, streamers, content creators, and anyone who uses an external microphone regularly. I believe this is becoming an increasingly important feature as wireless microphones become more common. Please consider adding independent microphone input and audio output selection in a future update. Thank you!
0
0
31
2d
BlietoothHFP to MFi hearingaids
Is it possible to port a bluetooth microphone (BluetoothHFP) to MFi hearing aids connected to an iPhone. When I try to do that the BluetoothHFP grabs input and output. When I pull the output away (BluetoothA2DP) it gives up the input.
6
0
549
2d
Supported diagnostics for a natural playback transition failure in a WKWebView hosting music.apple.com
We are evaluating an unreleased iPhone app that embeds the complete official music.apple.com page in WKWebView. The official website displays its own lyrics. Our app retains an independent website sign-in and local listening tools. A JavaScript bridge observes playback metadata and provides explicit transport commands through the page's existing MusicKit instance. It does not create a developer-owned MusicKit instance or extract lyrics, credentials, protected audio addresses, or DRM data. On an iPhone 14 Pro Max running iOS 27.2.1, a foreground track can play to its natural end, after which the queue position advances but the next track remains stopped at time zero. A filtered error observation reports MKError with reason NETWORK_ERROR. This identifies the SDK error category, not the underlying cause or request owner. The failure was observed before Safari inspection was attached as well as during inspection. Safari Web Inspector can connect to the app's main page. Our metadata-only observations include a fetch row with HTTP 200 and several subsequent fetch rows with no displayed HTTP status. We have not captured request addresses, headers, cookies, response bodies, HAR files, lyric text, or raw stacks. These rows cannot currently be linked to the SDK error. An earlier actual-zero-user-script control with the full native model and layout completed one foreground transition and continued playing the next track for 51 seconds. This is not a current matched comparison and does not identify a specific bridge operation. The complete bridge can also complete transitions in a separate Mac host. We have no deterministic minimal reproducer establishing an Apple or WebKit defect. The official-website branch leaves Media Session and audio-session ownership to the website and does not issue automatic next-track commands. No proposed global fetch wrapper has been installed, because adding Promise rejection handlers changes observable rejection behavior. We are seeking guidance before expanding instrumentation or requesting more real-device test time. Questions concerning this single transition failure: Is full music.apple.com playback in a third-party WKWebView, alongside observation and explicit transport through the page's existing MusicKit instance, a supported integration pattern? If not, which supported integration preserves a website account independent of the iPhone's Media & Purchases account? Which public diagnostics can distinguish cancellation from transport failure at this transition without exposing protected resource addresses, authorization values, response content, or raw stacks? Are there documented event-listener reentrancy, WKWebView lifecycle, or audio-session constraints that this host should follow when observing the existing player? Environment: iPhone 14 Pro Max; iOS 27.2.1; macOS 27.2.1; Xcode 27.2 (27B52). Experimental app version 0.65, build 67, not submitted. The existing public version is unchanged. Attachments: None. This draft contains no account identifiers, device identifiers, real-session logs, signing material, or source attachments.
1
0
171
4d
Let Audio Units read the project's whole tempo and signature map (FB25609451)
I've filed FB25609451 and I'm posting it here so other Audio Unit developers can add their own feedback if they need the same. What I'd love in Logic Pro (and any AU host) is what REAPER already does for plug-ins: you insert the plug-in and it immediately knows the whole song, every bar's time signature and tempo, without playing anything, and it updates right away when you change a signature or tempo. In REAPER this works through its host interface (IReaperHostApplication): the plug-in calls TimeMap_GetMeasureInfo to read the start, time signature and tempo of any bar. For Audio Units today, HostCallback_GetBeatAndTempo / HostCallback_GetMusicalTimeLocation (AUv2) and musicalContextBlock / transportStateBlock (AUv3) give the current tempo and time signature, but nothing about the other bars. So a time-aware plug-in has to hear the song played through once, or the user pastes the Signature List into it and redoes it after every edit. Suggestion, read-only: A way for an Audio Unit to read the project's tempo and signature map (Signature track and Tempo track), for example a host callback for AUv2 and a block or property on AUAudioUnit for AUv3, returning every event with its position, time signature and tempo. A notification when the user edits them. Also useful: the current bar number in musicalContextBlock / HostCallback_GetMusicalTimeLocation, and the playback rate during Varispeed. It would help click and count-in tools, arpeggiators and sequencers, tempo-synced effects, and notation or visual displays. If you need this too, filing your own feedback that mentions FB25609451 helps. Leo Margarit, Margarit Audio (OddClick)
0
0
49
5d
AUHAL input callback teardown: when is inputProcRefCon safe to release?
I am developing a macOS application that captures audio through an AUHAL input AudioUnit. The application registers an input callback using kAudioOutputUnitProperty_SetInputCallback. The callback receives an application-owned context through inputProcRefCon and obtains input audio using AudioUnitRender. The intended shutdown sequence is: Close application-level callback admission. Call AudioOutputUnitStop. Remove the callback using AudioUnitSetProperty with a zero-initialized AURenderCallbackStruct. Wait for callbacks already entered by the application to finish. Call AudioUnitUninitialize and AudioComponentInstanceDispose. Release the application-owned callback context. We use an atomic active-callback counter, covering all context accesses made by callbacks that have entered the counted region. However, this counter cannot account for a callback that Core Audio may have already dispatched but that has not yet executed its first instruction. My question: Is there a supported AUHAL teardown sequence that guarantees no previously scheduled, dispatched, or executing input callback can subsequently access inputProcRefCon? Specifically, does successful return from AudioOutputUnitStop, callback removal, AudioUnitUninitialize, or AudioComponentInstanceDispose provide such a guarantee? Is passing an AURenderCallbackStruct with both fields set to NULL a supported way to remove the input callback, and does removal have any documented completion semantics? If none of these operations provides the required guarantee independently, what is the recommended lifetime-management pattern for safely releasing the application-owned callback context? I have prepared a small standalone AUHAL test project that compiles the relevant Core Audio operations and passes deterministic offline lifecycle tests. Physical AUHAL activation and native callback teardown have not yet been executed, so I am not claiming a reproduced crash or verified native behavior. I am primarily seeking clarification of the supported Core Audio API contract, rather than a diagnosis of an observed crash. Thank you.
0
0
104
6d
AVSpeechSynthesizer pulls words out of thin air.
Hi, I'm working on a project that uses the AVSpeechSynthesizer and AVSpeechUtterance. I discovered by chance that the AVSpeechSynthesizer automatically completes some words instead of just outputting what it's supposed to. These are abbreviations for days of the week or months, but not all of them. I don't want either of them automatically completed, but only the specified text. The completion transcends languages. I have written a short example program for demonstration purposes. import SwiftUI import AVFoundation import Foundation let synthesizer: AVSpeechSynthesizer = AVSpeechSynthesizer() struct ContentView: View { var body: some View { VStack { Button { utter("mon") } label: { Text("mon") } .buttonStyle(.borderedProminent) Button { utter("tue") } label: { Text("tue") } .buttonStyle(.borderedProminent) Button { utter("thu") } label: { Text("thu") } .buttonStyle(.borderedProminent) Button { utter("feb") } label: { Text("feb") } .buttonStyle(.borderedProminent) Button { utter("feb", lang: "de-DE") } label: { Text("feb DE") } .buttonStyle(.borderedProminent) Button { utter("wed") } label: { Text("wed") } .buttonStyle(.borderedProminent) } .padding() } private func utter(_ text: String, lang: String = "en-US") { let utterance = AVSpeechUtterance(string: text) let voice = AVSpeechSynthesisVoice(language: lang) utterance.voice = voice synthesizer.speak(utterance) } } #Preview { ContentView() } Thank you Christian
5
0
1.9k
1w
AVAudioSession ioBufferDuration always default to 512 samples on MacOS27 no matter what is requested
Hello, I develop an audio processing demo app that has been working fine on iOS/iPad and on Mac ("developed for iPad") until I upgraded to MacOS27. The app is demonstrating low latency / high temporal resolution audio spectral analysis. I am using AVAudioSession, I request the audio IO buffer to be 32 samples. This still works fine on iOS/iPadOS, but it seems that when running on MacOS27 the buffer size always defaults to 512 samples no matter what I request (smaller or larger). This applies to the "Designed for iPad" version running on MacOS and any version running in the iPhone and iPad simulators on MacOS. The app seems to be working as intended on physical iPhones and iPads. I made a catalyst version and used AudioObjectSetPropertyData to force the IO bugger sizes to 32 samples on the system's active hardware device. This seems to succeed but still results in incorrect behaviour in my app; I am still diagnosing but I believe it's because this leaves AVAudioSession with incorrect data, and I am using AudioKit which relies on AVAudioSession (not confirmed). Regardless, is this change in behaviour intentional in MacOS27? Is there a way around it? The whole point of my app is to demonstrate efficient audio processing with very low latency - I need small input buffer size for that. Thank you for any help with this
2
0
442
1w
macOS 27.2 makes AUv3 plug-ins completely useless as soon as the user accesses a button with a native submenu.
https://youtu.be/HnNGgYZQBvYJust submitted a bug report FB25497228, and posting it here in hopes to get Apple engineers' attention because it's such a bad bug that macOS 27.2 introduced, and don't think it has a workaround unless Apple fixes it. This wasn't an issue in 27.0. Summary: Opening a native submenu in an out-of-process AUv3 plug-in and dismissing it without selecting anything will make the plug-in UI completely unresponsive to mouse input, and any animation will stop also. Only restart solves it. I can reproduce this with Mela in both Logic Pro and Ableton Live 12. I've attached a demo project to the bug report. It's Xcode Audio Unti extension template with a small modification, just added a button with a submenu. Mela can be found here if needed: https://apps.apple.com/us/app/mela-auv3-synth-fx-midi/id1668299428 Steps to reproduce with the attached project Load Mela as an AU instrument in Logic Pro or Ableton Live Open a factory patch. Open a module’s Actions menu (ellipsis buttons) Hover over Insert or Replace until the submenu appears, then dismiss it without selecting anything. Expected result Mela's UI continues to be interactive, e.g. parameters can be easily adjusted and modulation animations continue to work. Actual result No interaction can be performed, wether adjusting a prameter or dragging a module or pressing a button. Animations also stop. This bug renders my plug-in completely useless on macOS as it makes heavy use of submenus. Please fix it! Here's a link to the video demo of the issue: https://youtu.be/HnNGgYZQBvY
1
0
142
1w
iOS 27 Public Beta 6: CallKit UI remains after ending two outgoing calls in modified Apple Speakerbox sample
Summary On an iPhone 17 running iOS 27 Public Beta 6, the system CallKit UI remains visible after ending two outgoing calls individually. This is reproducible using Apple's “Making and Receiving VoIP Calls with CallKit” sample (Speakerbox), with only two minimal changes to support multiple simultaneous calls. After both calls end, Speakerbox displays “No Calls”, but the system CallKit UI still shows 111, and the Dynamic Island retains a call-related status. Environment Device: iPhone 17 (physical device) OS: iOS 27 Public Beta 6 App: Apple's Speakerbox sample with the modifications below Backend: None. The sample's original simulated calls are used; no external VoIP SDK is involved. Changes to the original sample Only two files were modified. 1. ProviderDelegate.swift Added: providerConfiguration.maximumCallGroups = 3 The original setting remains unchanged: providerConfiguration.maximumCallsPerCallGroup = 1 2. SpeakerboxCallManager.swift Before starting a new outgoing call, hold existing calls that are not already on hold. The hold actions and new start action are submitted in the same transaction: let transaction = CXTransaction() for call in calls where !call.isOnHold { transaction.addAction( CXSetHeldCallAction(call: call.uuid, onHold: true) ) } transaction.addAction(startCallAction) requestTransaction(transaction) No changes were made to the simulated connection timing, startedConnectingAt/connectedAt reporting, CXEndCallAction handling, call removal, or audio handling. The original end-call handler finds the call by UUID, calls endSpeakerboxCall(), fulfills the end action, and removes the call from the app's list. Steps to reproduce Launch the modified Speakerbox sample. Create an outgoing call to 111. Create a second outgoing call to 222. The first call is placed on hold. Wait until Speakerbox shows 111 — On Hold and 222 — Active, with call durations displayed. In Speakerbox, select 222 - End Call. Select 111 - End Call. Confirm that Speakerbox displays “No Calls”. Open the app switcher and inspect the system CallKit UI. Also inspect the Dynamic Island. Open the remaining system call UI and tap End. Expected result After both calls end, the system should no longer display an ongoing call in the CallKit UI or Dynamic Island. Actual result Speakerbox displays “No Calls”. The system CallKit UI continues to display the first call, 111. The Dynamic Island continues to show a call-related “Hold” status. Tapping End in the remaining system call UI does not clear the residual state during the recording. The issue occurs even after waiting for both simulated calls to connect before ending them. This is the system-provided CallKit UI, not an active-call view implemented by the sample app. Reproduction video Recording: 4b03d30e507b874d9c19c0523632e9d6.mp4 00:24–00:26 (approximately): Both calls are present; 111 is on hold and 222 is active. 00:26–00:31 (approximately): End 222, then end 111. Speakerbox displays “No Calls”. 00:31–01:03 (approximately): The system CallKit UI and Dynamic Island retain the call state despite further attempts to end it. Questions Is this an iOS 27 regression, or is additional lifecycle handling required when extending Speakerbox to support multiple calls? Please advise what additional diagnostics would help investigate why the system UI retains an ongoing-call state after both calls have been ended using the sample's original end-call handling. Source and video downloads https://github.com/willbin/callkit-ios27-repro/releases/tag/repro-v1
6
0
1.4k
1w
AVAudioIONode presentationLatency discrepancy
The documentation for -[AVAudioIONode presentationLatency] states that it accounts for both kAudioDevicePropertyLatency and kAudioStreamPropertyLatency. On my MBP the following values are reported for the speakers in the output scope: kAudioDevicePropertyLatency: 11 kAudioStreamPropertyLatency: 390 At the device's nominal sample rate of 44.1 kHz that equates to approximately 9.09 msec. However, the output node on a running AVAudioEngine reports presentationLatency as approximately 0.25 msec. This appears to be the device latency only. Is it necessary to compensate for the stream latency manually despite the documentation?
0
0
231
1w
Is MusicKit playback allowed on Meta Quest, and can a preview play as background music for a profile panorama?
I'm building Welink, a mixed-reality app for Meta Quest where people who opt in see each other's profile cards in person. Users connect Apple Music through MusicKit JS, and their card shows recently played songs with Apple Music links, attribution and tap-to-play 30-second previews (streamed, never saved). A few policy questions: Is MusicKit (MusicKit for Android or MusicKit JS in the headset browser) supported and permitted on Meta Quest's Android-based OS for full playback by Apple Music subscribers? Guideline 5.2.5 says previews can't be used for "entertainment value." Would playing a user-chosen song (preview, or the full song for subscribers) when a viewer steps into that person's 360° panorama be allowed, given that the viewer starts it, can stop it, and full attribution is shown? Is it fine to show a user's recently played songs (metadata and cover art) to people they meet, when the user turned that on and can turn it off at any time? Can previews and cover art be cached on-device for about 24 hours for offline demos? Can the Apple Music panel appear in a public demo video, following the Apple Music Identity Guidelines? Any pointers to the right docs or team would be appreciated. Thanks!
0
0
138
1w
MusicKit JS authorize() returns AUTHORIZATION_ERROR "Unauthorized" (no Music User Token) — catalog works
Hi all — hoping someone has hit this. Our MusicKit JS (v3) web app calls authorize() and it rejects with AUTHORIZATION_ERROR / "Unauthorized" (no Music User Token is issued), even though the same developer token returns HTTP 200 for catalog requests (/v1/catalog/us/search) — so the token itself is valid. Setup (all verified): MusicKit JS v3 from js-cdn.music.apple.com/musickit/v3/musickit.js; MusicKit.configure() succeeds, storefront resolves to "us". Developer token is ES256, valid, and includes the origin claim for our exact site origin. Our MusicKit key is enabled for Media Services and tied to a Media ID that has MusicKit enabled. Referrer-Policy: strict-origin-when-cross-origin. Repro: sign in with an active Apple Music subscriber and tap Allow -> authorize() rejects. Reproduces for two subscriber Apple IDs, in Chrome and Edge, with third-party cookies allowed. Error object: { name: "AUTHORIZATION_ERROR", message: "Unauthorized", isMKError: true }, thrown inside musickit.js during the authorize() flow. The served MusicKit build is a prerelease (3.2526.0-prerelease.x). We've ruled out the usual suspects on our side (origin claim, referrer policy, token validity, API version, cookies, and portal provisioning). Has anyone resolved this, or is there an additional server-side enablement needed for a team/Media ID to issue web user tokens? Also filed as FB23587284. Thanks!
4
0
1.1k
1w
macOS 27 regression: HDMI + DisplayPort Multi-Output audio fails with Core Audio error 1937010544
Hi, I believe I have found a Core Audio regression in macOS 27. This exact hardware configuration worked correctly on macOS 26, and the issue started immediately after upgrading to macOS 27. Hardware: Mac mini M2 2 × MSI MP243X monitors Monitor 1 connected via HDMI Monitor 2 connected via DisplayPort Both monitors expose a 2-channel audio output to macOS On macOS 26: I used Audio MIDI Setup to create a Multi-Output Device containing both MSI MP243X audio outputs. Audio played simultaneously through both monitors without any issues. On macOS 27: Both audio devices are still detected correctly and both work perfectly when selected individually: HDMI only: WORKS DisplayPort only: WORKS HDMI + DisplayPort simultaneously: FAILS This is reproducible even immediately after a full reboot. When I create a new Multi-Output Device containing both MSI outputs and select it as the system output, audio playback cannot start. For example: afplay /System/Library/Sounds/Glass.aiff returns: Error: AudioQueueStart failed ('stop') I captured Core Audio debug logs while reproducing the problem. Core Audio initially reports that StartIO succeeded, but the physical display audio device then fails to establish a timeline. Relevant log messages include: HALS_IOEngine2::_StartIO succeeded Device 3669B540-0000-0000-2623-0104A5351D78 is not running. could not establish a timeline after waiting 10000000 microseconds stopping with error 1937010544 StartIOThread: the IO thread failed to start, Error: 1937010544 I have already tested the following: Both devices at 44.1 kHz / 24-bit Both devices at 48 kHz / 16-bit Recreating the Multi-Output Device from scratch Using the HDMI device as clock source Using the DisplayPort device as clock source Different drift correction configurations Restarting coreaudiod Full macOS reboot Testing both outputs individually immediately after reboot Alternative multi-device audio routing software The behavior is always reproducible. Each MSI MP243X audio endpoint works correctly by itself. The failure only occurs when both HDMI and DisplayPort audio endpoints are opened simultaneously. system_profiler correctly detects both devices at 48 kHz: MSI MP243X Output Channels: 2 Current Sample Rate: 48000 Transport: HDMI MSI MP243X Output Channels: 2 Current Sample Rate: 48000 Transport: DisplayPort This does not appear to be a problem with either monitor or connection individually, because both outputs continue to work independently. The important difference is the macOS version: macOS 26: HDMI + DisplayPort Multi-Output = WORKING macOS 27: HDMI only = WORKING DisplayPort only = WORKING HDMI + DisplayPort Multi-Output = FAILS with AudioQueueStart 'stop' / Core Audio error 1937010544 No hardware, monitor, cable, or connection changes were made between the working macOS 26 configuration and the macOS 27 configuration. This therefore appears to be a regression in the Core Audio/display-audio path introduced with macOS 27 when two display audio endpoints (HDMI + DisplayPort) are opened simultaneously. Has anyone else been able to reproduce this on macOS 27, particularly on Apple Silicon? I can provide the complete coreaudiod logs and run additional Core Audio diagnostics if needed. Additional isolation tests: Both monitors can remain physically connected to the Mac at the same time. With both HDMI and DisplayPort connected: Selecting the HDMI MSI MP243X as the only output works correctly. Selecting the DisplayPort MSI MP243X as the only output works correctly. Opening both audio endpoints simultaneously causes the failure. Therefore, simply having both displays connected does not trigger the issue. The failure appears specifically when Core Audio attempts to run both display-audio endpoints at the same time. I also tested this using third-party multi-device audio routing implementations instead of relying only on Audio MIDI Setup. The same behavior occurs: either MSI MP243X can be opened and used independently, but attempting to route audio to both devices simultaneously fails. This suggests that the issue is below the Multi-Output Device configuration itself and may be related to starting/synchronizing two display-audio IO engines simultaneously. During the failure, the Core Audio log sequence is particularly interesting: The physical device is detected normally. The audio format is configured successfully. Core Audio reports that StartIO succeeded. The device subsequently reports that it is not running. Core Audio waits 10 seconds for a timeline. Timeline establishment fails. The IO thread terminates with error 1937010544 ('stop'). One captured conversion path was: 2 ch, 48000 Hz, Float32, interleaved -> 2 ch, 48000 Hz, Int16, interleaved so both sides were operating at 48 kHz during this test. Another important observation is that restarting coreaudiod does not resolve the simultaneous-output problem. A complete reboot also does not resolve it: immediately after a fresh boot, before launching any third-party audio-routing software, both physical outputs work individually but a newly created Multi-Output Device still reproduces the failure. This makes the issue highly reproducible on this machine. Minimal reproduction: Boot macOS 27 with both MSI MP243X monitors connected. Open Audio MIDI Setup. Verify both MSI MP243X devices work independently. Set both devices to 48,000 Hz. Create a new Multi-Output Device. Select both MSI MP243X devices. Select the new Multi-Output Device as the system audio output. Run: afplay /System/Library/Sounds/Glass.aiff Expected result: The sound is reproduced simultaneously through both monitors, as it was on macOS 26. Actual result: Error: AudioQueueStart failed ('stop') Switching the system output back to either individual MSI MP243X restores audio playback. The problem can then be reproduced again by selecting the Multi-Output Device.
7
10
2.1k
1w
MusicTokenRequestError.unknown for some authorized, subscribed users
A small number of our users can’t connect Apple Music because DefaultMusicTokenProvider.userToken(for:options:) always throws MusicTokenRequestError.unknown. Most users connect without problems and we can’t reproduce the failure on any of our own accounts. Flow: let status = await MusicAuthorization.request() // .authorized let subscription = try await MusicSubscription.current // canPlayCatalogContent == true let provider = DefaultMusicTokenProvider() let developerToken = try await provider.developerToken(options: []) // succeeds let userToken = try await provider.userToken(for: developerToken, options: []) // throws .unknown On failure we retry once with options: .ignoreCache for both the developer token and the user token. The retry fails in the same way. What we see on affected devices: MusicAuthorization.currentStatus == .authorized MusicSubscription.current.canPlayCatalogContent == true The developer token is returned successfully. userToken(for:) throws MusicTokenRequestError.unknown. The NSError has code 0 and the localized description “Unknown error”, and there is no NSUnderlyingError. The ignoreCache retry fails after anywhere from about 10 ms to 2.5 s, so it sometimes appears to go out to the network. For a given account it fails every time: across days, app versions, and after reinstalling both our app and the Music app. Apple Music plays normally in the Music app for these users. Ruled out so far: iOS version: seen on iOS 26.6.1 and 27.2. Device model: seen on iPhone 13 and iPhone 14 Plus. Region: seen in Hungary, Italy, Romania and Japan. Subscription type: one affected user has an individual trial. A family member with her own individual trial on a separate account connects without problems. Family Sharing in general: family plan members connect fine for us. Questions: What conditions cause userToken(for:) to return .unknown instead of a specific case such as .userNotSignedIn, .permissionDenied or .privacyAcknowledgementRequired? Is there a way to get more detail about the cause? Is there a step affected users can take to fix it?
3
1
626
1w
Technical question about iPhone Voice Memos microphone processing for music recording
Hey, I’m an independent musician and producer experimenting with recording acoustic instruments using the iPhone’s built-in microphones and Voice Memos. I’ve found that the recordings have a particularly pleasing tonal balance and dynamic character, sometimes more so than recordings made with my conventional studio microphones. I’m interested in understanding what processing, if any, contributes to this sound, so I can experiment with recreating it in my recording workflow. Specifically, I’d love to know: Does Voice Memos apply automatic gain control, dynamic range compression, limiting or other dynamics processing? If so what is the ratio, attack, threshold, and release. Is any frequency-dependent EQ, noise reduction or microphone correction applied? How does stereo recording combine signals from the different built-in microphones? Is the processing fixed, or does it adapt to the incoming signal? Is there a way to capture a minimally processed recording from the same microphones for comparison? I’m not expecting proprietary algorithms or exact settings, but even a general technical description of the signal path would be incredibly helpful. Thanks very much for your time! Jethro Karis
0
0
263
1w
Can a third-party app capture microphone audio while a cellular call is active?
We're building a companion app that people use during a regular cellular call (VoIP isn't an option for this call). The user wants to briefly mute the call, speak a short question into our app, then unmute. Is there any supported way for a third-party app to record from the microphone during an active cellular call, even if the call is muted or on hold? Does Siri with App Intents accept voice input during an active call? If so, is Siri's spoken response heard by the other person on the call? Can a watchOS app record audio while the paired iPhone is on a call? If none of these work, what approach do you recommend?
1
0
165
2w
MusicKit: Multichannel Dolby Atmos Limited to Stereo Output - Is This Intended Behavior?
I'm experiencing a significant limitation with MusicKit's Dolby Atmos implementation on macOS and would appreciate clarification on whether this is intended behavior or if there are solutions available. When streaming Dolby Atmos content through MusicKit's ApplicationMusicPlayer, the output is limited to 2-channel stereo, even when: Audio MIDI Setup is configured for 7.1.4 (12-channel) output The same tracks play in full multichannel through the native Apple Music app Dolby Atmos is set to "Automatic" in Apple Music preferences Please let me know if there is anyway to enable this. If not, is this documented anywhere? Thanks!
2
3
422
2w
Logic Pro: third-party effect AU not processed on an empty audio track — request for the opt-in some vendors have (aumf/Acnt/EINX)
We have a product where Logic Pro's "process effects only when needed" optimisation blocks the core use case, and we would like to ask for the opt-in that some third-party plug-ins already have. Filed as FB25426476. The plug-in ACENTAUR (Audio Unit v2, type aumf, subtype Acnt, manufacturer EINX; EINAUX, Germany) is an AI music generator that lives in the Audio FX slot of an audio track. It renders its own material from the host's tempo, time signature and transport, so the natural workflow is: empty audio track, insert ACENTAUR, press Play, generate. There is no region and no input signal. The plug-in is the source. What happens On an empty audio track Logic never calls the render callback, so the plug-in never receives the host clock through the HostCallbacks: tempo and transport stay unknown and our UI stays locked. Everything else works: the same build renders in the instrument slot, in other hosts, and on a Logic audio track as soon as a region or live input is present. Inserting Logic's Test Oscillator (bypassed) in the same strip also makes it work, which shows the strip is simply not scheduled. Logic engineering described this optimisation on coreaudio-api in February 2024 (https://www.mail-archive.com/coreaudio-api@lists.apple.com/msg01859.html). What we observed A number of third-party effects (Celemony Melodyne, Antares, Synchro Arts, Waves and others) are processed on an empty track. We verified on Logic Pro 12.4 (6707) / macOS 26.6.2 that this is tied to the plug-in's identity plus a private global property Logic queries when the plug-in is loaded: with a listed manufacturer code and that property answered, our plug-in is rendered immediately on an empty track and the host clock arrives. We are of course not going to ship under another vendor's identity; this was a one-off local test to make sure we understand the mechanism before asking. Log excerpt from that test (plug-in log, out-of-process hosting, empty audio track, 48 kHz): 09:19:38 host asked property 101000 → 0 09:19:38 host asked property 101001 → 1 09:19:38 Initialize → 0 · in 2 ch @ 48000 Hz · out 2 ch 09:19:39 First render after 0s · input callback · tempo 120.00 Same build under our own manufacturer code EINX: "No render after 30s · host clock tempo+transport · flags 0". The request Could ACENTAUR (aumf / Acnt / EINX) be added to the set of plug-ins Logic keeps processing without input? The plug-in already answers the property in question; the only missing piece is on Logic's side. If there is a different or public mechanism you would prefer third parties to adopt (a component tag, an AU property, an AUv3 property), we will implement it right away. Happy to provide a build, a demo project and whatever else helps. Christoph Kahse, EINAUX
0
0
110
2w
iPhone 18 Pro: AVAudioEngine input tap returns all zeros after the Always-On Display turns off overnight
My app records audio overnight in the background: UIBackgroundModes audio, AVAudioSession category playAndRecord with mode measurement, and an AVAudioEngine with a tap on inputNode. On a customer's iPhone 18 Pro running iOS 27.2.1 (24B156), the tap starts delivering buffers of exact zeros about two hours into the night. It keeps delivering zeros until she picks the phone up in the morning. Her sysdiagnose shows the same sequence on three consecutive nights: powerd logs SmartPowerNap: enterSmartPowerNap. The backlight state changes from inactiveOnLive to suppressed, with the explanation "alwaysOn suppressed - SmartPowerNap". The panel reaches 0 nits about two seconds later. Under a second after that, every sample in the recording is 0. PowerLog's per-app mic RMS reads -240 from that point on. Hours later powerd logs Exiting SmartPowerNap. Session end reason MotionDetected. The display wakes and mic input returns within a second. Timing on the third night: 23:55:57.234 Smart Power Nap entry 23:55:59.14 panel at 0 nits 23:55:59.8 first zero sample in the recording 05:46:58.032 Smart Power Nap exit about 05:46:59 mic input returns During the silent hours: The app receives no interruption, route change, media services reset or AVAudioEngineConfigurationChange notification. The engine keeps running and the tap keeps firing on schedule. audiomxd logs no deactivation for the session and holds its power assertions throughout. The system never sleeps. I can't reproduce it. My iPhone 16 Pro on the same build logs "SmartPowerNap: feature not enabled or predictor not initialized" on every lock, so it never takes this path. Questions: Does iOS gate mic input by design while Smart Power Nap suppresses the Always-On Display? Can an app keep recording through it, or get a notification that input has stopped? Does restarting AVAudioEngine while the session stays active restore input in this state? I can detect a run of exact zeros and restart the engine, but I have no device to test it on. Has anyone else with iPhone 18 Pro users seen this? I filed FB25663635 with the sysdiagnose attached.
Replies
0
Boosts
0
Views
73
Activity
1d
Playback volume stays low with PlayAndRecord/VoiceChat and remains low for a few seconds after restoring SoloAmbient (iPhone Air, iOS 26)
Hello, I'm developing a speech recognition app with Unity 6000.0.58f2. The app also has other content besides speech recognition, so I switch the AVAudioSession only when the user needs to speak, and restore it once recording is no longer needed. On iPhone Air (iOS 26), I'm seeing the following issues: While the session is set to PlayAndRecord with VoiceChat mode, playback volume is consistently lower than normal. After restoring the session to SoloAmbient, playback volume stays low for a few seconds before returning to normal. I'd like to avoid this volume drop. Any advice or recommended approach would be greatly appreciated. [Implementation] I use a native iOS plugin (.mm file) called from Unity. When the user starts speaking, I enable recording with: success = [session setCategory:AVAudioSessionCategoryPlayAndRecord mode:AVAudioSessionModeVoiceChat options:AVAudioSessionCategoryOptionDefaultToSpeaker | AVAudioSessionCategoryOptionAllowBluetooth error:&error];After speaking ends, I restore the session with: success = [session setCategory:AVAudioSessionCategorySoloAmbient error:&error]; success = [session setMode:AVAudioSessionModeDefault error:&error];[Environment] Device: iPhone Air iOS: 26.[x] Unity: 6000.0.58f2 Reproduces on other devices / iOS versions: [Yes / No — e.g. iPhone 15 (iOS 18.x) does not reproduce] Output route: built-in speaker [/ Bluetooth: ___ ] [Questions] Is the lower playback volume in VoiceChat mode expected behavior due to voice processing (echo cancellation / ducking)? If so, is there a recommended way to keep playback volume at a normal level while recording for speech recognition (e.g. using a different mode such as Default or Measurement, or adjusting voice processing ducking)? What is the recommended sequence for restoring the session so that playback volume returns to normal immediately? For example, should I call setActive:NO (with NotifyOthersOnDeactivation) before changing the category, or set the category and mode together in a single call? Is there any behavior change in iOS 26 or anything specific to iPhone Air that could cause this? Thank you in advance.
Replies
0
Boosts
0
Views
155
Activity
2d
Feature Request: Allow Independent Bluetooth Microphone Input and Speaker Output on iPhone
I hope iphone can consider adding this feature in a future update. We are now living in the era of content creation and social media. With the growing popularity of Bluetooth wireless microphones, more and more people are using external microphones for gaming, live streaming, video calls, and content creation. However, there is a frustrating limitation: when connecting a Bluetooth microphone, the phone may automatically route audio output to the same Bluetooth device. This means users cannot freely use an external microphone for input while keeping the phone's built-in speakers for audio output. What we really need is the ability to independently select audio input and output devices at the system level, just like on a PC. For example: Audio input: DJI wireless microphone Audio output: Phone's built-in speakers This would be extremely useful for mobile gamers, streamers, content creators, and anyone who uses an external microphone regularly. I believe this is becoming an increasingly important feature as wireless microphones become more common. Please consider adding independent microphone input and audio output selection in a future update. Thank you!
Replies
0
Boosts
0
Views
31
Activity
2d
BlietoothHFP to MFi hearingaids
Is it possible to port a bluetooth microphone (BluetoothHFP) to MFi hearing aids connected to an iPhone. When I try to do that the BluetoothHFP grabs input and output. When I pull the output away (BluetoothA2DP) it gives up the input.
Replies
6
Boosts
0
Views
549
Activity
2d
Supported diagnostics for a natural playback transition failure in a WKWebView hosting music.apple.com
We are evaluating an unreleased iPhone app that embeds the complete official music.apple.com page in WKWebView. The official website displays its own lyrics. Our app retains an independent website sign-in and local listening tools. A JavaScript bridge observes playback metadata and provides explicit transport commands through the page's existing MusicKit instance. It does not create a developer-owned MusicKit instance or extract lyrics, credentials, protected audio addresses, or DRM data. On an iPhone 14 Pro Max running iOS 27.2.1, a foreground track can play to its natural end, after which the queue position advances but the next track remains stopped at time zero. A filtered error observation reports MKError with reason NETWORK_ERROR. This identifies the SDK error category, not the underlying cause or request owner. The failure was observed before Safari inspection was attached as well as during inspection. Safari Web Inspector can connect to the app's main page. Our metadata-only observations include a fetch row with HTTP 200 and several subsequent fetch rows with no displayed HTTP status. We have not captured request addresses, headers, cookies, response bodies, HAR files, lyric text, or raw stacks. These rows cannot currently be linked to the SDK error. An earlier actual-zero-user-script control with the full native model and layout completed one foreground transition and continued playing the next track for 51 seconds. This is not a current matched comparison and does not identify a specific bridge operation. The complete bridge can also complete transitions in a separate Mac host. We have no deterministic minimal reproducer establishing an Apple or WebKit defect. The official-website branch leaves Media Session and audio-session ownership to the website and does not issue automatic next-track commands. No proposed global fetch wrapper has been installed, because adding Promise rejection handlers changes observable rejection behavior. We are seeking guidance before expanding instrumentation or requesting more real-device test time. Questions concerning this single transition failure: Is full music.apple.com playback in a third-party WKWebView, alongside observation and explicit transport through the page's existing MusicKit instance, a supported integration pattern? If not, which supported integration preserves a website account independent of the iPhone's Media & Purchases account? Which public diagnostics can distinguish cancellation from transport failure at this transition without exposing protected resource addresses, authorization values, response content, or raw stacks? Are there documented event-listener reentrancy, WKWebView lifecycle, or audio-session constraints that this host should follow when observing the existing player? Environment: iPhone 14 Pro Max; iOS 27.2.1; macOS 27.2.1; Xcode 27.2 (27B52). Experimental app version 0.65, build 67, not submitted. The existing public version is unchanged. Attachments: None. This draft contains no account identifiers, device identifiers, real-session logs, signing material, or source attachments.
Replies
1
Boosts
0
Views
171
Activity
4d
Let Audio Units read the project's whole tempo and signature map (FB25609451)
I've filed FB25609451 and I'm posting it here so other Audio Unit developers can add their own feedback if they need the same. What I'd love in Logic Pro (and any AU host) is what REAPER already does for plug-ins: you insert the plug-in and it immediately knows the whole song, every bar's time signature and tempo, without playing anything, and it updates right away when you change a signature or tempo. In REAPER this works through its host interface (IReaperHostApplication): the plug-in calls TimeMap_GetMeasureInfo to read the start, time signature and tempo of any bar. For Audio Units today, HostCallback_GetBeatAndTempo / HostCallback_GetMusicalTimeLocation (AUv2) and musicalContextBlock / transportStateBlock (AUv3) give the current tempo and time signature, but nothing about the other bars. So a time-aware plug-in has to hear the song played through once, or the user pastes the Signature List into it and redoes it after every edit. Suggestion, read-only: A way for an Audio Unit to read the project's tempo and signature map (Signature track and Tempo track), for example a host callback for AUv2 and a block or property on AUAudioUnit for AUv3, returning every event with its position, time signature and tempo. A notification when the user edits them. Also useful: the current bar number in musicalContextBlock / HostCallback_GetMusicalTimeLocation, and the playback rate during Varispeed. It would help click and count-in tools, arpeggiators and sequencers, tempo-synced effects, and notation or visual displays. If you need this too, filing your own feedback that mentions FB25609451 helps. Leo Margarit, Margarit Audio (OddClick)
Replies
0
Boosts
0
Views
49
Activity
5d
AUHAL input callback teardown: when is inputProcRefCon safe to release?
I am developing a macOS application that captures audio through an AUHAL input AudioUnit. The application registers an input callback using kAudioOutputUnitProperty_SetInputCallback. The callback receives an application-owned context through inputProcRefCon and obtains input audio using AudioUnitRender. The intended shutdown sequence is: Close application-level callback admission. Call AudioOutputUnitStop. Remove the callback using AudioUnitSetProperty with a zero-initialized AURenderCallbackStruct. Wait for callbacks already entered by the application to finish. Call AudioUnitUninitialize and AudioComponentInstanceDispose. Release the application-owned callback context. We use an atomic active-callback counter, covering all context accesses made by callbacks that have entered the counted region. However, this counter cannot account for a callback that Core Audio may have already dispatched but that has not yet executed its first instruction. My question: Is there a supported AUHAL teardown sequence that guarantees no previously scheduled, dispatched, or executing input callback can subsequently access inputProcRefCon? Specifically, does successful return from AudioOutputUnitStop, callback removal, AudioUnitUninitialize, or AudioComponentInstanceDispose provide such a guarantee? Is passing an AURenderCallbackStruct with both fields set to NULL a supported way to remove the input callback, and does removal have any documented completion semantics? If none of these operations provides the required guarantee independently, what is the recommended lifetime-management pattern for safely releasing the application-owned callback context? I have prepared a small standalone AUHAL test project that compiles the relevant Core Audio operations and passes deterministic offline lifecycle tests. Physical AUHAL activation and native callback teardown have not yet been executed, so I am not claiming a reproduced crash or verified native behavior. I am primarily seeking clarification of the supported Core Audio API contract, rather than a diagnosis of an observed crash. Thank you.
Replies
0
Boosts
0
Views
104
Activity
6d
AVSpeechSynthesizer pulls words out of thin air.
Hi, I'm working on a project that uses the AVSpeechSynthesizer and AVSpeechUtterance. I discovered by chance that the AVSpeechSynthesizer automatically completes some words instead of just outputting what it's supposed to. These are abbreviations for days of the week or months, but not all of them. I don't want either of them automatically completed, but only the specified text. The completion transcends languages. I have written a short example program for demonstration purposes. import SwiftUI import AVFoundation import Foundation let synthesizer: AVSpeechSynthesizer = AVSpeechSynthesizer() struct ContentView: View { var body: some View { VStack { Button { utter("mon") } label: { Text("mon") } .buttonStyle(.borderedProminent) Button { utter("tue") } label: { Text("tue") } .buttonStyle(.borderedProminent) Button { utter("thu") } label: { Text("thu") } .buttonStyle(.borderedProminent) Button { utter("feb") } label: { Text("feb") } .buttonStyle(.borderedProminent) Button { utter("feb", lang: "de-DE") } label: { Text("feb DE") } .buttonStyle(.borderedProminent) Button { utter("wed") } label: { Text("wed") } .buttonStyle(.borderedProminent) } .padding() } private func utter(_ text: String, lang: String = "en-US") { let utterance = AVSpeechUtterance(string: text) let voice = AVSpeechSynthesisVoice(language: lang) utterance.voice = voice synthesizer.speak(utterance) } } #Preview { ContentView() } Thank you Christian
Replies
5
Boosts
0
Views
1.9k
Activity
1w
AVAudioSession ioBufferDuration always default to 512 samples on MacOS27 no matter what is requested
Hello, I develop an audio processing demo app that has been working fine on iOS/iPad and on Mac ("developed for iPad") until I upgraded to MacOS27. The app is demonstrating low latency / high temporal resolution audio spectral analysis. I am using AVAudioSession, I request the audio IO buffer to be 32 samples. This still works fine on iOS/iPadOS, but it seems that when running on MacOS27 the buffer size always defaults to 512 samples no matter what I request (smaller or larger). This applies to the "Designed for iPad" version running on MacOS and any version running in the iPhone and iPad simulators on MacOS. The app seems to be working as intended on physical iPhones and iPads. I made a catalyst version and used AudioObjectSetPropertyData to force the IO bugger sizes to 32 samples on the system's active hardware device. This seems to succeed but still results in incorrect behaviour in my app; I am still diagnosing but I believe it's because this leaves AVAudioSession with incorrect data, and I am using AudioKit which relies on AVAudioSession (not confirmed). Regardless, is this change in behaviour intentional in MacOS27? Is there a way around it? The whole point of my app is to demonstrate efficient audio processing with very low latency - I need small input buffer size for that. Thank you for any help with this
Replies
2
Boosts
0
Views
442
Activity
1w
macOS 27.2 makes AUv3 plug-ins completely useless as soon as the user accesses a button with a native submenu.
https://youtu.be/HnNGgYZQBvYJust submitted a bug report FB25497228, and posting it here in hopes to get Apple engineers' attention because it's such a bad bug that macOS 27.2 introduced, and don't think it has a workaround unless Apple fixes it. This wasn't an issue in 27.0. Summary: Opening a native submenu in an out-of-process AUv3 plug-in and dismissing it without selecting anything will make the plug-in UI completely unresponsive to mouse input, and any animation will stop also. Only restart solves it. I can reproduce this with Mela in both Logic Pro and Ableton Live 12. I've attached a demo project to the bug report. It's Xcode Audio Unti extension template with a small modification, just added a button with a submenu. Mela can be found here if needed: https://apps.apple.com/us/app/mela-auv3-synth-fx-midi/id1668299428 Steps to reproduce with the attached project Load Mela as an AU instrument in Logic Pro or Ableton Live Open a factory patch. Open a module’s Actions menu (ellipsis buttons) Hover over Insert or Replace until the submenu appears, then dismiss it without selecting anything. Expected result Mela's UI continues to be interactive, e.g. parameters can be easily adjusted and modulation animations continue to work. Actual result No interaction can be performed, wether adjusting a prameter or dragging a module or pressing a button. Animations also stop. This bug renders my plug-in completely useless on macOS as it makes heavy use of submenus. Please fix it! Here's a link to the video demo of the issue: https://youtu.be/HnNGgYZQBvY
Replies
1
Boosts
0
Views
142
Activity
1w
iOS 27 Public Beta 6: CallKit UI remains after ending two outgoing calls in modified Apple Speakerbox sample
Summary On an iPhone 17 running iOS 27 Public Beta 6, the system CallKit UI remains visible after ending two outgoing calls individually. This is reproducible using Apple's “Making and Receiving VoIP Calls with CallKit” sample (Speakerbox), with only two minimal changes to support multiple simultaneous calls. After both calls end, Speakerbox displays “No Calls”, but the system CallKit UI still shows 111, and the Dynamic Island retains a call-related status. Environment Device: iPhone 17 (physical device) OS: iOS 27 Public Beta 6 App: Apple's Speakerbox sample with the modifications below Backend: None. The sample's original simulated calls are used; no external VoIP SDK is involved. Changes to the original sample Only two files were modified. 1. ProviderDelegate.swift Added: providerConfiguration.maximumCallGroups = 3 The original setting remains unchanged: providerConfiguration.maximumCallsPerCallGroup = 1 2. SpeakerboxCallManager.swift Before starting a new outgoing call, hold existing calls that are not already on hold. The hold actions and new start action are submitted in the same transaction: let transaction = CXTransaction() for call in calls where !call.isOnHold { transaction.addAction( CXSetHeldCallAction(call: call.uuid, onHold: true) ) } transaction.addAction(startCallAction) requestTransaction(transaction) No changes were made to the simulated connection timing, startedConnectingAt/connectedAt reporting, CXEndCallAction handling, call removal, or audio handling. The original end-call handler finds the call by UUID, calls endSpeakerboxCall(), fulfills the end action, and removes the call from the app's list. Steps to reproduce Launch the modified Speakerbox sample. Create an outgoing call to 111. Create a second outgoing call to 222. The first call is placed on hold. Wait until Speakerbox shows 111 — On Hold and 222 — Active, with call durations displayed. In Speakerbox, select 222 - End Call. Select 111 - End Call. Confirm that Speakerbox displays “No Calls”. Open the app switcher and inspect the system CallKit UI. Also inspect the Dynamic Island. Open the remaining system call UI and tap End. Expected result After both calls end, the system should no longer display an ongoing call in the CallKit UI or Dynamic Island. Actual result Speakerbox displays “No Calls”. The system CallKit UI continues to display the first call, 111. The Dynamic Island continues to show a call-related “Hold” status. Tapping End in the remaining system call UI does not clear the residual state during the recording. The issue occurs even after waiting for both simulated calls to connect before ending them. This is the system-provided CallKit UI, not an active-call view implemented by the sample app. Reproduction video Recording: 4b03d30e507b874d9c19c0523632e9d6.mp4 00:24–00:26 (approximately): Both calls are present; 111 is on hold and 222 is active. 00:26–00:31 (approximately): End 222, then end 111. Speakerbox displays “No Calls”. 00:31–01:03 (approximately): The system CallKit UI and Dynamic Island retain the call state despite further attempts to end it. Questions Is this an iOS 27 regression, or is additional lifecycle handling required when extending Speakerbox to support multiple calls? Please advise what additional diagnostics would help investigate why the system UI retains an ongoing-call state after both calls have been ended using the sample's original end-call handling. Source and video downloads https://github.com/willbin/callkit-ios27-repro/releases/tag/repro-v1
Replies
6
Boosts
0
Views
1.4k
Activity
1w
AVAudioIONode presentationLatency discrepancy
The documentation for -[AVAudioIONode presentationLatency] states that it accounts for both kAudioDevicePropertyLatency and kAudioStreamPropertyLatency. On my MBP the following values are reported for the speakers in the output scope: kAudioDevicePropertyLatency: 11 kAudioStreamPropertyLatency: 390 At the device's nominal sample rate of 44.1 kHz that equates to approximately 9.09 msec. However, the output node on a running AVAudioEngine reports presentationLatency as approximately 0.25 msec. This appears to be the device latency only. Is it necessary to compensate for the stream latency manually despite the documentation?
Replies
0
Boosts
0
Views
231
Activity
1w
Is MusicKit playback allowed on Meta Quest, and can a preview play as background music for a profile panorama?
I'm building Welink, a mixed-reality app for Meta Quest where people who opt in see each other's profile cards in person. Users connect Apple Music through MusicKit JS, and their card shows recently played songs with Apple Music links, attribution and tap-to-play 30-second previews (streamed, never saved). A few policy questions: Is MusicKit (MusicKit for Android or MusicKit JS in the headset browser) supported and permitted on Meta Quest's Android-based OS for full playback by Apple Music subscribers? Guideline 5.2.5 says previews can't be used for "entertainment value." Would playing a user-chosen song (preview, or the full song for subscribers) when a viewer steps into that person's 360° panorama be allowed, given that the viewer starts it, can stop it, and full attribution is shown? Is it fine to show a user's recently played songs (metadata and cover art) to people they meet, when the user turned that on and can turn it off at any time? Can previews and cover art be cached on-device for about 24 hours for offline demos? Can the Apple Music panel appear in a public demo video, following the Apple Music Identity Guidelines? Any pointers to the right docs or team would be appreciated. Thanks!
Replies
0
Boosts
0
Views
138
Activity
1w
MusicKit JS authorize() returns AUTHORIZATION_ERROR "Unauthorized" (no Music User Token) — catalog works
Hi all — hoping someone has hit this. Our MusicKit JS (v3) web app calls authorize() and it rejects with AUTHORIZATION_ERROR / "Unauthorized" (no Music User Token is issued), even though the same developer token returns HTTP 200 for catalog requests (/v1/catalog/us/search) — so the token itself is valid. Setup (all verified): MusicKit JS v3 from js-cdn.music.apple.com/musickit/v3/musickit.js; MusicKit.configure() succeeds, storefront resolves to "us". Developer token is ES256, valid, and includes the origin claim for our exact site origin. Our MusicKit key is enabled for Media Services and tied to a Media ID that has MusicKit enabled. Referrer-Policy: strict-origin-when-cross-origin. Repro: sign in with an active Apple Music subscriber and tap Allow -> authorize() rejects. Reproduces for two subscriber Apple IDs, in Chrome and Edge, with third-party cookies allowed. Error object: { name: "AUTHORIZATION_ERROR", message: "Unauthorized", isMKError: true }, thrown inside musickit.js during the authorize() flow. The served MusicKit build is a prerelease (3.2526.0-prerelease.x). We've ruled out the usual suspects on our side (origin claim, referrer policy, token validity, API version, cookies, and portal provisioning). Has anyone resolved this, or is there an additional server-side enablement needed for a team/Media ID to issue web user tokens? Also filed as FB23587284. Thanks!
Replies
4
Boosts
0
Views
1.1k
Activity
1w
macOS 27 regression: HDMI + DisplayPort Multi-Output audio fails with Core Audio error 1937010544
Hi, I believe I have found a Core Audio regression in macOS 27. This exact hardware configuration worked correctly on macOS 26, and the issue started immediately after upgrading to macOS 27. Hardware: Mac mini M2 2 × MSI MP243X monitors Monitor 1 connected via HDMI Monitor 2 connected via DisplayPort Both monitors expose a 2-channel audio output to macOS On macOS 26: I used Audio MIDI Setup to create a Multi-Output Device containing both MSI MP243X audio outputs. Audio played simultaneously through both monitors without any issues. On macOS 27: Both audio devices are still detected correctly and both work perfectly when selected individually: HDMI only: WORKS DisplayPort only: WORKS HDMI + DisplayPort simultaneously: FAILS This is reproducible even immediately after a full reboot. When I create a new Multi-Output Device containing both MSI outputs and select it as the system output, audio playback cannot start. For example: afplay /System/Library/Sounds/Glass.aiff returns: Error: AudioQueueStart failed ('stop') I captured Core Audio debug logs while reproducing the problem. Core Audio initially reports that StartIO succeeded, but the physical display audio device then fails to establish a timeline. Relevant log messages include: HALS_IOEngine2::_StartIO succeeded Device 3669B540-0000-0000-2623-0104A5351D78 is not running. could not establish a timeline after waiting 10000000 microseconds stopping with error 1937010544 StartIOThread: the IO thread failed to start, Error: 1937010544 I have already tested the following: Both devices at 44.1 kHz / 24-bit Both devices at 48 kHz / 16-bit Recreating the Multi-Output Device from scratch Using the HDMI device as clock source Using the DisplayPort device as clock source Different drift correction configurations Restarting coreaudiod Full macOS reboot Testing both outputs individually immediately after reboot Alternative multi-device audio routing software The behavior is always reproducible. Each MSI MP243X audio endpoint works correctly by itself. The failure only occurs when both HDMI and DisplayPort audio endpoints are opened simultaneously. system_profiler correctly detects both devices at 48 kHz: MSI MP243X Output Channels: 2 Current Sample Rate: 48000 Transport: HDMI MSI MP243X Output Channels: 2 Current Sample Rate: 48000 Transport: DisplayPort This does not appear to be a problem with either monitor or connection individually, because both outputs continue to work independently. The important difference is the macOS version: macOS 26: HDMI + DisplayPort Multi-Output = WORKING macOS 27: HDMI only = WORKING DisplayPort only = WORKING HDMI + DisplayPort Multi-Output = FAILS with AudioQueueStart 'stop' / Core Audio error 1937010544 No hardware, monitor, cable, or connection changes were made between the working macOS 26 configuration and the macOS 27 configuration. This therefore appears to be a regression in the Core Audio/display-audio path introduced with macOS 27 when two display audio endpoints (HDMI + DisplayPort) are opened simultaneously. Has anyone else been able to reproduce this on macOS 27, particularly on Apple Silicon? I can provide the complete coreaudiod logs and run additional Core Audio diagnostics if needed. Additional isolation tests: Both monitors can remain physically connected to the Mac at the same time. With both HDMI and DisplayPort connected: Selecting the HDMI MSI MP243X as the only output works correctly. Selecting the DisplayPort MSI MP243X as the only output works correctly. Opening both audio endpoints simultaneously causes the failure. Therefore, simply having both displays connected does not trigger the issue. The failure appears specifically when Core Audio attempts to run both display-audio endpoints at the same time. I also tested this using third-party multi-device audio routing implementations instead of relying only on Audio MIDI Setup. The same behavior occurs: either MSI MP243X can be opened and used independently, but attempting to route audio to both devices simultaneously fails. This suggests that the issue is below the Multi-Output Device configuration itself and may be related to starting/synchronizing two display-audio IO engines simultaneously. During the failure, the Core Audio log sequence is particularly interesting: The physical device is detected normally. The audio format is configured successfully. Core Audio reports that StartIO succeeded. The device subsequently reports that it is not running. Core Audio waits 10 seconds for a timeline. Timeline establishment fails. The IO thread terminates with error 1937010544 ('stop'). One captured conversion path was: 2 ch, 48000 Hz, Float32, interleaved -> 2 ch, 48000 Hz, Int16, interleaved so both sides were operating at 48 kHz during this test. Another important observation is that restarting coreaudiod does not resolve the simultaneous-output problem. A complete reboot also does not resolve it: immediately after a fresh boot, before launching any third-party audio-routing software, both physical outputs work individually but a newly created Multi-Output Device still reproduces the failure. This makes the issue highly reproducible on this machine. Minimal reproduction: Boot macOS 27 with both MSI MP243X monitors connected. Open Audio MIDI Setup. Verify both MSI MP243X devices work independently. Set both devices to 48,000 Hz. Create a new Multi-Output Device. Select both MSI MP243X devices. Select the new Multi-Output Device as the system audio output. Run: afplay /System/Library/Sounds/Glass.aiff Expected result: The sound is reproduced simultaneously through both monitors, as it was on macOS 26. Actual result: Error: AudioQueueStart failed ('stop') Switching the system output back to either individual MSI MP243X restores audio playback. The problem can then be reproduced again by selecting the Multi-Output Device.
Replies
7
Boosts
10
Views
2.1k
Activity
1w
MusicTokenRequestError.unknown for some authorized, subscribed users
A small number of our users can’t connect Apple Music because DefaultMusicTokenProvider.userToken(for:options:) always throws MusicTokenRequestError.unknown. Most users connect without problems and we can’t reproduce the failure on any of our own accounts. Flow: let status = await MusicAuthorization.request() // .authorized let subscription = try await MusicSubscription.current // canPlayCatalogContent == true let provider = DefaultMusicTokenProvider() let developerToken = try await provider.developerToken(options: []) // succeeds let userToken = try await provider.userToken(for: developerToken, options: []) // throws .unknown On failure we retry once with options: .ignoreCache for both the developer token and the user token. The retry fails in the same way. What we see on affected devices: MusicAuthorization.currentStatus == .authorized MusicSubscription.current.canPlayCatalogContent == true The developer token is returned successfully. userToken(for:) throws MusicTokenRequestError.unknown. The NSError has code 0 and the localized description “Unknown error”, and there is no NSUnderlyingError. The ignoreCache retry fails after anywhere from about 10 ms to 2.5 s, so it sometimes appears to go out to the network. For a given account it fails every time: across days, app versions, and after reinstalling both our app and the Music app. Apple Music plays normally in the Music app for these users. Ruled out so far: iOS version: seen on iOS 26.6.1 and 27.2. Device model: seen on iPhone 13 and iPhone 14 Plus. Region: seen in Hungary, Italy, Romania and Japan. Subscription type: one affected user has an individual trial. A family member with her own individual trial on a separate account connects without problems. Family Sharing in general: family plan members connect fine for us. Questions: What conditions cause userToken(for:) to return .unknown instead of a specific case such as .userNotSignedIn, .permissionDenied or .privacyAcknowledgementRequired? Is there a way to get more detail about the cause? Is there a step affected users can take to fix it?
Replies
3
Boosts
1
Views
626
Activity
1w
Technical question about iPhone Voice Memos microphone processing for music recording
Hey, I’m an independent musician and producer experimenting with recording acoustic instruments using the iPhone’s built-in microphones and Voice Memos. I’ve found that the recordings have a particularly pleasing tonal balance and dynamic character, sometimes more so than recordings made with my conventional studio microphones. I’m interested in understanding what processing, if any, contributes to this sound, so I can experiment with recreating it in my recording workflow. Specifically, I’d love to know: Does Voice Memos apply automatic gain control, dynamic range compression, limiting or other dynamics processing? If so what is the ratio, attack, threshold, and release. Is any frequency-dependent EQ, noise reduction or microphone correction applied? How does stereo recording combine signals from the different built-in microphones? Is the processing fixed, or does it adapt to the incoming signal? Is there a way to capture a minimally processed recording from the same microphones for comparison? I’m not expecting proprietary algorithms or exact settings, but even a general technical description of the signal path would be incredibly helpful. Thanks very much for your time! Jethro Karis
Replies
0
Boosts
0
Views
263
Activity
1w
Can a third-party app capture microphone audio while a cellular call is active?
We're building a companion app that people use during a regular cellular call (VoIP isn't an option for this call). The user wants to briefly mute the call, speak a short question into our app, then unmute. Is there any supported way for a third-party app to record from the microphone during an active cellular call, even if the call is muted or on hold? Does Siri with App Intents accept voice input during an active call? If so, is Siri's spoken response heard by the other person on the call? Can a watchOS app record audio while the paired iPhone is on a call? If none of these work, what approach do you recommend?
Replies
1
Boosts
0
Views
165
Activity
2w
MusicKit: Multichannel Dolby Atmos Limited to Stereo Output - Is This Intended Behavior?
I'm experiencing a significant limitation with MusicKit's Dolby Atmos implementation on macOS and would appreciate clarification on whether this is intended behavior or if there are solutions available. When streaming Dolby Atmos content through MusicKit's ApplicationMusicPlayer, the output is limited to 2-channel stereo, even when: Audio MIDI Setup is configured for 7.1.4 (12-channel) output The same tracks play in full multichannel through the native Apple Music app Dolby Atmos is set to "Automatic" in Apple Music preferences Please let me know if there is anyway to enable this. If not, is this documented anywhere? Thanks!
Replies
2
Boosts
3
Views
422
Activity
2w
Logic Pro: third-party effect AU not processed on an empty audio track — request for the opt-in some vendors have (aumf/Acnt/EINX)
We have a product where Logic Pro's "process effects only when needed" optimisation blocks the core use case, and we would like to ask for the opt-in that some third-party plug-ins already have. Filed as FB25426476. The plug-in ACENTAUR (Audio Unit v2, type aumf, subtype Acnt, manufacturer EINX; EINAUX, Germany) is an AI music generator that lives in the Audio FX slot of an audio track. It renders its own material from the host's tempo, time signature and transport, so the natural workflow is: empty audio track, insert ACENTAUR, press Play, generate. There is no region and no input signal. The plug-in is the source. What happens On an empty audio track Logic never calls the render callback, so the plug-in never receives the host clock through the HostCallbacks: tempo and transport stay unknown and our UI stays locked. Everything else works: the same build renders in the instrument slot, in other hosts, and on a Logic audio track as soon as a region or live input is present. Inserting Logic's Test Oscillator (bypassed) in the same strip also makes it work, which shows the strip is simply not scheduled. Logic engineering described this optimisation on coreaudio-api in February 2024 (https://www.mail-archive.com/coreaudio-api@lists.apple.com/msg01859.html). What we observed A number of third-party effects (Celemony Melodyne, Antares, Synchro Arts, Waves and others) are processed on an empty track. We verified on Logic Pro 12.4 (6707) / macOS 26.6.2 that this is tied to the plug-in's identity plus a private global property Logic queries when the plug-in is loaded: with a listed manufacturer code and that property answered, our plug-in is rendered immediately on an empty track and the host clock arrives. We are of course not going to ship under another vendor's identity; this was a one-off local test to make sure we understand the mechanism before asking. Log excerpt from that test (plug-in log, out-of-process hosting, empty audio track, 48 kHz): 09:19:38 host asked property 101000 → 0 09:19:38 host asked property 101001 → 1 09:19:38 Initialize → 0 · in 2 ch @ 48000 Hz · out 2 ch 09:19:39 First render after 0s · input callback · tempo 120.00 Same build under our own manufacturer code EINX: "No render after 30s · host clock tempo+transport · flags 0". The request Could ACENTAUR (aumf / Acnt / EINX) be added to the set of plug-ins Logic keeps processing without input? The plug-in already answers the property in question; the only missing piece is on Logic's side. If there is a different or public mechanism you would prefer third parties to adopt (a component tag, an AU property, an AUv3 property), we will implement it right away. Happy to provide a build, a demo project and whatever else helps. Christoph Kahse, EINAUX
Replies
0
Boosts
0
Views
110
Activity
2w