AirFlash supports existing two-device stereo pairs and low-latency playback from Windows. Stream Windows desktop audio to HomePod through a native AirPlay 2 sender, with support for HomePod OS 27.
AirFlash is an AirPlay 2 alternative to TuneBlade (AirPlay 1), designed to bring Windows audio streaming to HomePods running newer versions of HomePod software.
Download the latest release from GitHub Releases, run
AirFlash-X.Y.Z.msi, and launch AirFlash from the installed shortcut. Apps and Features provides repair and uninstall.
The minimum target is Windows x64, Windows 10 1809 / build 17763 (LTSC 2019), including LTSC 2021, 22H2 and Windows 11. Windows 10 qualification is pending affected-user testing. See Windows compatibility and startup diagnostics for the candidate status and the PowerShell 5.1 diagnostic collector.
AirFlash is distributed as a Windows x64 executable and standalone MSI. Windows may show a SmartScreen warning because release binaries are not code-signed. Use More info and Run anyway only when the downloaded files are from a source you trust.
For audio capture, we recommend Settings > Audio capture > Specific audio endpoint. If you do not have an unused audio endpoint available, install a virtual endpoint with Virtual Audio Cable (VAC Lite) or a similar tool, and route the audio you want to stream to that endpoint.
- Streams Windows system audio to HomePod over AirPlay 2.
- Supports synchronized playback on existing two-device HomePod stereo pairs.
- Adds experimental two-receiver software stereo groups, independent hardware volume, per-receiver PCM/ALAC and UDP/TCP profiles, and 5 ms continuity transitions.
- Offers automatic device discovery, PIN pairing, and manual device setup.
- Lets you filter device discovery by selecting a network adapter in Settings.
- Includes selectable latency modes, tray operation, mute restoration, automatic reconnection, and connection diagnostics.
- Provides a ten-band equalizer with presets and live preview in Settings.
- Captures a selected application or playback/recording endpoints, including virtual audio devices.
- Releases AirPlay sessions after configurable silence and reconnects when audio returns.
- Available as a portable Windows x64 app with English, Simplified Chinese, Traditional Chinese, and German UI.
See audio continuity and schema v4 for setup, protocol details, simulation results and remaining hardware acceptance. Samsung, Mac and mixed-brand playback and acoustic synchronization have not been qualified by this implementation.
- Dolby Atmos support.
- Connect the PC and both HomePods to the same reachable local network.
- Start AirFlash and wait for the receiver list to populate.
- Select the complete
HomePod stereo · 2/2entry and press play. - Complete PIN pairing in Settings > Receivers if the receiver requests it.
If pairing fails with
pairing rejected (TLV error [2]), open the Home app, set speaker access to Everyone, and try again. - Choose the capture endpoint and latency profile in Settings when the defaults are not suitable.
Stream formats are configured in Settings > Receivers for each receiver:
codec, sample rate, UDP/TCP transport, and clock strategy. Choices are filtered by
the receiver's /info capabilities without starting playback or pairing. Automatic
prefers a supported 44.1 kHz profile, then 48 kHz. Its label shows the expected
selection before playback and the negotiated selection while playing; unapplied
edits continue to show predictions. Offline or undetected receivers offer Automatic
and retain existing explicit settings as pending verification. Native stereo pairs
use their members' common capabilities. Software stereo groups control both members'
common sample rate and default to 44.1 kHz.
The control panel's Global latency mode changes the shared default. While the current connecting, playing, or standby receiver uses an independent latency mode, including either software stereo member, the control is disabled and points to Settings > Receivers. Stopping playback restores it; the global default remains editable in Settings.
New installations and reset settings default to Normal · 200 ms. The selectable modes are Real-time, Normal, Buffered, and Custom. Existing settings that used the retired adaptive mode load as Normal; other fixed modes and custom values are preserved.
In Settings > AirFlash streaming, the 120/200/500 ms presets and 0–2000 ms custom target also offer an independent Force fixed latency switch. With it off, the existing recovery behavior remains. With it on, AirFlash does not automatically increase latency or move the playback timeline: short stalls may use brief silence and clear stale capture audio, while larger stalls reconnect at the same target within the configured retry limit.
Receiver settings can inherit the global policy, keep recovery, or force a fixed target. Forced software stereo requires every member to use the same forced target; conflicts are reported before playback without changing the group or its left/right synchronization offsets. Applying a target or policy change during playback rebuilds the session.
A forced target below the receiver's reported minimum shows a warning and is still attempted; the receiver may reject playback or remain silent. The control panel distinguishes the requested target, engine plan, and receiver minimum. These scheduling values are not measured end-to-end audio latency. Without forcing, the existing negotiation of receiver latency remains in effect. Latency mode and UDP/TCP transport are independent settings.
Stopping playback or cancelling a connection immediately shows Disconnecting…. The playback button becomes available again after the old session is cleaned up. AirFlash drains final engine output during shutdown and gives the engine a shared two-second stop budget before terminating an unresponsive engine process.
Settings > General > Reset and restart asks once before clearing all settings
and receiver pairing. AirFlash stops playback and discovery, restores local mute,
disables startup at login, deletes config.json and native-credentials, and restarts
with defaults. Unsaved edits are discarded; logs and engine caches are retained.
Schema v4 removes the global sample rate: v1–v3 inherited rates become Automatic,
while explicit receiver and software group rates are preserved.
In Settings > Audio capture, Specific audio endpoint lists active playback and recording devices with their type. For Voicemeeter, select the recording endpoint corresponding to its B1/B2/B3 bus (for example, Voicemeeter Output) and enable that bus for the desired audio in Voicemeeter. Playback devices use loopback; recording devices capture their input directly. Automatic local mute applies only to playback devices and never mutes a recording input.
Specific application captures the selected application and its child processes; browsers include the complete process tree, not individual tabs. This native mode requires Windows build 20348 or later. Older Windows versions retain endpoint capture. The application selection is saved by executable path and follows a restart when exactly one matching instance is running. If several independent instances match, select an instance again. A missing application never falls back to system audio.
To hear the target application only on HomePod while games and notifications stay on your headphones, install a virtual cable such as VB-CABLE and use Open Windows volume mixer to route only that application's output to CABLE Input. Select the application, or use Specific audio endpoint → CABLE Output · Recording. Disable Listen to this device and any local monitoring. Application capture never mutes your entire playback device. The manual Windows routing remains after AirFlash stops; change it back when local playback is needed. Virtual routing cannot guarantee capture of DRM-protected content, and capture, resampling, and stream-format conversion do not guarantee bit-perfect output.
Enable AirFlash streaming → Standby to release the complete AirPlay session after 5–300 seconds of silence (default duration: 10 seconds; disabled by default). Capture continues locally, and detected audio automatically starts a new session. Reconnection resumes current audio; its opening or short notification sounds may be missed. HomePod manages its own power state after the session is released.
Receiver discovery runs while the control panel or Settings > Receivers is visible, and remains active for software stereo topology changes during playback. Otherwise it pauses when both are hidden. Previous results remain available; pausing discovery does not interrupt playback. If automatic connection is enabled globally or for a receiver, AirFlash also discovers devices for ten seconds at startup. Open a receiver list to resume discovery afterwards.
AirFlash intentionally uses fresh application storage under %APPDATA%/AirFlash
and %LOCALAPPDATA%/AirFlash. It does not read configuration, credentials, or
engine caches created by the previous product name; users must configure and pair
again after installing AirFlash.
Use Settings > About > Check for updates to check for a stable release and open its download page. Checks run only when requested.
The real-time profile targets 120 ms; target and local transport timings are not measurements of end-to-end acoustic latency.
For PTP exchange did not become ready, see PTP startup troubleshooting
for clock diagnostics and the correct engine process and UDP firewall rules.
See Release workflow for version reservation and publishing.
See Icon maintenance for the vector source, generated resources, and high-DPI verification.
- C# / .NET 10 / WPF for the Windows UI, tray integration, settings, and discovery
- Rust for WASAPI capture, AirPlay 2 sessions, HAP, PTP, RTP, and audio transport
- Windows DNS-SD for receiver discovery
- WASAPI loopback for system audio capture
- PCM and ALAC with encrypted RTP for HomePod playback
- uv, pytest, and Ruff for the finite validation tools and Python checks
The published application is self-contained and statically links the native MSVC runtime. It does not require Python, a separately installed .NET runtime, or a separately installed VC++ Redistributable.
Development requires Windows 10/11 x64, the .NET 10 SDK, the Rust MSVC toolchain,
and Visual Studio C++/Windows SDK components. Python 3.12+ and uv are used for
validation scripts only.
uv sync --locked
pwsh scripts/build-native.ps1 -Check
pwsh scripts/dotnet.ps1 restore AirFlash.sln --locked-mode
pwsh scripts/dotnet.ps1 run --project AirFlash.App/AirFlash.App.csproj
pwsh scripts/build-dev.ps1The development build produces:
dist/dev/AirFlash.exedist/dev/AirFlash-X.Y.Z-dev.<commit>.msi
The development package is labeled from the latest stable vX.Y.Z tag and the
current commit. It does not reserve a release version or push a tag. Uninstall an
existing same-version AirFlash package before installing a development MSI.
Use pwsh scripts/build-release.ps1 only when performing a release; it requires a
clean release branch, reserves the next numeric version, and writes dist/.
Run the MSI directly. It installs the self-contained executable and provides optional desktop and Start menu shortcuts.
AirFlash uses the Windows DNS-SD API and does not require Bonjour or install a background service.
The Rust engine is embedded in the WPF executable and extracted to a content-addressed cache under %LOCALAPPDATA%/AirFlash/engine/<hash> at runtime.
pwsh scripts/dotnet.ps1 test AirFlash.sln
uv run pytest
uv run ruff check .
pwsh scripts/build-native.ps1 -CheckThe ignored Rust soak test uses localhost UDP receivers only. Real HomePod tests are intentionally limited to the repository's low-volume, five-second procedure.
See Discovery stability for the isolated playback retention regression check and repair validation notes.
Keep protocol behavior, pairing boundaries, and measurement limits explicit in changes. Pull requests are welcome!
