Skip to content

WIP: Route playback through VPIO for echo cancellation on iOS #264

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
kavimathur wants to merge 6 commits into
base: main
Choose a base branch
from
Open

WIP: Route playback through VPIO for echo cancellation on iOS #264

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
858b568
WIP: Route playback through VPIO for echo cancellation on iOS
Feb 13, 2026
e5aa09a
Add design doc for VPIO echo cancellation approach
Feb 13, 2026
c6fc665
Fix iOS realtime self-interruption with VPIO echo guard
Feb 14, 2026
a6c6801
Clarify actor annotation caveat on Core Audio callbacks
Feb 15, 2026
8ea8098
Annotate iOS-only guards as iOS-first pending cross-platform validation
Feb 15, 2026
fb2480c
Expose Realtime input audio noise reduction config
Feb 16, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
55 changes: 43 additions & 12 deletions Sources/AIProxy/AudioController.swift
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,16 +18,16 @@ import AVFoundation
/// We use either AVAudioEngine or AudioToolbox for mic data, depending on the platform and whether headphones are attached.
/// The following arrangement provides for the best user experience:
///
/// +----------+---------------+------------------+
/// | Platform | Headphones | Audio API |
/// +----------+---------------+------------------+
/// | macOS | Yes | AudioEngine |
/// | macOS | No | AudioToolbox |
/// | iOS | Yes | AudioEngine |
/// | iOS | No | AudioToolbox |
/// | watchOS | Yes | AudioEngine |
/// | watchOS | No | AudioEngine |
/// +----------+---------------+------------------+
/// +----------+---------------+--------------------------------------+
/// | Platform | Headphones | Audio API |
/// +----------+---------------+--------------------------------------+
/// | macOS | Yes | AudioEngine |
/// | macOS | No | AudioToolbox |
/// | iOS | Yes | AudioEngine |
/// | iOS | No | AudioToolbox + manual rendering AEC |
/// | watchOS | Yes | AudioEngine |
/// | watchOS | No | AudioEngine |
/// +----------+---------------+--------------------------------------+
///
@AIProxyActor public final class AudioController {
public enum Mode {
Expand All @@ -42,7 +42,14 @@ import AVFoundation

public init(modes: [Mode]) async throws {
self.modes = modes
#if os(iOS)
#if os(iOS) // iOS-first guard: non-iOS behavior has not been validated for this path yet.
let needsManualRendering = modes.contains(.record) && modes.contains(.playback)
&& !AIProxyUtils.headphonesConnected
#else
let needsManualRendering = false
#endif

#if os(iOS) // iOS-first guard: non-iOS behavior has not been validated for this path yet.
// This is not respected if `setVoiceProcessingEnabled(true)` is used :/
// Instead, I've added my own accumulator.
// try? AVAudioSession.sharedInstance().setPreferredIOBufferDuration(0.1)
Expand All @@ -61,11 +68,27 @@ import AVFoundation

self.audioEngine = AVAudioEngine()

#if os(iOS) // iOS-first guard: non-iOS behavior has not been validated for this path yet.
if needsManualRendering {
let renderFormat = AVAudioFormat(
commonFormat: .pcmFormatFloat32,
sampleRate: 44100,
channels: 1,
interleaved: true
)!
try audioEngine.enableManualRenderingMode(
.realtime,
format: renderFormat,
maximumFrameCount: 4096
)
}
#endif

if modes.contains(.record) {
#if os(macOS) || os(iOS)
self.microphonePCMSampleVendor = AIProxyUtils.headphonesConnected
? try MicrophonePCMSampleVendorAE(audioEngine: self.audioEngine)
: MicrophonePCMSampleVendorAT()
: MicrophonePCMSampleVendorAT(audioEngine: self.audioEngine)
#else
self.microphonePCMSampleVendor = try MicrophonePCMSampleVendorAE(audioEngine: self.audioEngine)
#endif
Expand All @@ -81,9 +104,17 @@ import AVFoundation
// Nesting `start` in a Task is necessary on watchOS.
// There is some sort of race, and letting the runloop tick seems to "fix" it.
// If I call `prepare` and `start` in serial succession, then there is no playback on watchOS (sometimes).
#if os(iOS) // iOS-first guard: non-iOS behavior has not been validated for this path yet.
try self.audioEngine.start()
#elseif os(watchOS)
Task {
try self.audioEngine.start()
}
#else
Task {
try self.audioEngine.start()
}
#endif
}

deinit {
Expand Down
Loading