Vibepollo is a fork of Nonary/vibepollo (itself a fork of ClassicOldSong/Apollo) that adds encrypted microphone passthrough from a Vibelight client to the Windows host during a game streaming session. Built with AI assistance (Claude Code) by a UX designer — hence the name.
When you stream games from your Windows PC to your Steam Deck, your microphone stays on your PC — the client mic is ignored. Vibepollo fixes that.
It pairs with Vibelight on the client side. Vibelight captures mic audio on the Steam Deck, Opus-encodes it, and streams it back to Vibepollo, which renders it as a virtual microphone input that Windows applications can use for voice chat, Discord, games, and anything else.
- Encrypted mic passthrough — mic audio rides the AES-GCM encrypted control stream (SS_ENC_CONTROL_V2). Plaintext mic is refused. No unencrypted audio on the network.
- Steam Streaming Microphone — decoded mic audio is written directly to the Steam Streaming Microphone render endpoint. No third-party driver required — uses the Steam audio driver already installed with Steam. Vibepollo normalizes the endpoint format to 2ch/32-bit/48kHz before WASAPI initialization.
- Opus audio — 64kbps mono, FEC enabled, 20ms frames. Low latency, excellent voice quality, robust to packet loss.
- Per-session decoder — each streaming session gets its own Opus decoder and WASAPI render client. Multiple concurrent sessions do not corrupt each other's audio.
- Session stats and diagnostics — logs packet count, PLC events, decode errors, audio level, and latency every 30 seconds during active mic sessions.
- All Apollo/Nonary features — virtual display, HDR, Playnite integration, and everything else from the upstream forks.
Windows PC (host):
- Windows 10 or 11
- Steam running on the PC (provides the Steam Streaming Microphone audio driver)
- GPU: NVIDIA (NVENC), AMD, or Intel (QuickSync)
Steam Deck (client):
- Vibelight — the companion Moonlight fork with client-side mic capture
- Download
Vibepollo-1.15.11-Setup.exefrom Releases - Run the installer — it handles everything automatically:
- Installs to
C:\Program Files\Vibepollo\ - Configures Windows Firewall rules (TCP 47984/47989/47990, UDP 47998-48010)
- Sets up autostart on login via Task Scheduler (30-second delay for audio init)
- Creates Start Menu shortcut
- Opens the web UI on first run for initial setup
- Installs to
- Install Vibelight on your Steam Deck
- Pair Vibelight with Vibepollo via the web UI at
https://localhost:47990
Requirements: MSYS2 with UCRT64 toolchain, CMake, Ninja
git clone https://github.com/xenstalker02/Vibepollo.git
cd Vibepollo
mkdir build && cd build
cmake .. -G Ninja -DCMAKE_BUILD_TYPE=Release
ninja sunshineRun from the build directory (working directory must be the build folder):
Start-Process -FilePath '.\sunshine.exe' -WorkingDirectory (Get-Location).PathTo rebuild the installer after making changes:
& "C:\Program Files (x86)\Inno Setup 6\ISCC.exe" installer\vibepollo.issVibepollo is configured via sunshine.conf in the config directory.
See sunshine.conf.template for all available options.
Key mic passthrough options:
| Option | Default | Description |
|---|---|---|
mic_sink |
Speakers (Steam Streaming Microphone) |
Render endpoint for mic passthrough. Vibepollo writes decoded Opus audio here. Steam must be running for this endpoint to be present. |
mic_capture_device |
Microphone (Steam Streaming Microphone) |
Capture device Vibepollo switches Windows default input to at session start, so Discord (on Default) picks up the client mic. Restored to previous default on session end. |
mic_buffer_packets |
2 |
Jitter buffer prebuffer depth in Opus packets (1 packet = 20ms). Default 2 = 40ms prebuffer. Range 1–16. Increase if audio drops; decrease if latency is too high. |
The web UI at https://localhost:47990 provides a graphical interface for most settings.
Set Discord (and other voice apps) input to Default — that's all. At session start, Vibepollo switches the Windows default capture device to Microphone (Steam Streaming Microphone), so Discord automatically picks up the client mic. At session end, the default capture restores to your previous default device. No manual switching needed.
Games that don't pick up the device switch automatically: Some games cache the Windows default audio device at launch and ignore changes made mid-session. If a game doesn't detect the client mic, go into the game's audio settings and manually select Microphone (Steam Streaming Microphone) as the microphone input — this is a one-time per-game setting. Vibepollo re-applies the device switch at 2, 5, 10, and 20 seconds after session start to catch games that init audio after process launch.
Mic packets are encrypted as part of the AES-GCM encrypted Moonlight control stream
(SS_ENC_CONTROL_V2). Vibepollo refuses to render mic audio from clients that did not
negotiate an encrypted control stream. Plaintext mic passthrough is never permitted.
Steam Deck mic
→ SDL2 capture (48kHz, 16-bit, stereo or mono)
→ (L+R)/2 downmix to mono
→ Opus encode (64kbps mono, FEC, 20ms frames)
→ 4-byte header prepended (channel/flags)
→ AES-GCM encrypted control stream (SS_ENC_CONTROL_V2)
→ Vibepollo receives 0x3003 packets
→ Jitter buffer (40ms prebuffer, 2 packets default)
→ Opus decode with PLC on packet loss
→ WASAPI render → Steam Streaming Microphone render endpoint
→ Vibepollo switches Windows default capture to Microphone (Steam Streaming Microphone)
→ Discord (set to Default) picks up client mic automatically
→ Session end: default capture restores to previous default device
| Platform | Status |
|---|---|
| Windows 10 / 11 (host) | Supported |
| Linux | Not supported |
| macOS | Not supported |
| Project | Description |
|---|---|
| Vibelight | Companion Moonlight fork with client mic capture |
| logabell/Apollo | Parallel server-side mic passthrough implementation |
| logabell/moonlight-qt-mic | Parallel client-side mic implementation — not compatible with Vibelight (uses LiSendMicrophoneOpusDataEx on a separate UDP port; Vibelight uses 0x3003 on the encrypted control stream) |
| Nonary/vibepollo | Our direct upstream |
| ClassicOldSong/Apollo | Apollo upstream |
| LizardByte/Sunshine | Sunshine upstream |
See ACKNOWLEDGMENTS.md for full credits.
Mic passthrough was developed in parallel with logabell. We compared implementations and adopted Opus encoder tuning (64kbps mono, reduced from Logan's original 96kbps for VOIP tuning; FEC, VBR, complexity 10, FRAMESIZE_20_MS) and the deadline-based send pacer from that work.