Skip to Content
User ManualVRChat Audio Routing

VRChat Audio Routing

This guide explains how to use AudioIn, AudioOut, RemoteAudioIn, and RemoteAudioOut with VB-Cable A+B so OpenNeuro can hear VRChat voice chat and talk back to other players.

Cable Mapping

Use the two virtual cables like this:

  • CABLE-A carries AI speech into VRChat’s microphone
  • CABLE-B carries VRChat’s game audio into OpenNeuro

The recommended signal flow is:

OpenNeuro AudioOut -> CABLE-A Input -> CABLE-A Output -> VRChat microphone
Windows default output / VRChat output -> CABLE-B Input -> CABLE-B Output -> OpenNeuro AudioIn

Important VRChat Note

VRChat lets you change the microphone device, but its output follows the Windows system playback device. That means:

  • Set the VRChat microphone to CABLE-A Output
  • Set the Windows default playback device to CABLE-B Input

If you still want to hear the game locally, enable Listen to this device on CABLE-B Output and route it to your real headphones or speakers.

Local Setup

Use this when OpenNeuro and VRChat are running on the same machine.

OpenNeuro Components

Configure the graph like this:

  • AudioIn
    • device: CABLE-B Output
    • sample_rate: 48000
    • channels: 2
    • frame_ms: 20
  • AudioOut
    • device: CABLE-A Input
    • sample_rate: 48000
    • channels: 1

VRChat / Windows

  • In VRChat, set Microphone to CABLE-A Output
  • In Windows, set the default playback device to CABLE-B Input
  • Optional: enable Listen to this device on CABLE-B Output so you can monitor game audio through your real headset

Remote Setup

Use this when OpenNeuro and VRChat are running on different machines.

On the Game Machine

Install VB-Cable A+B and start the bridge server:

cd backend
uv run python -m src.lib.audio.remote_audio_server

By default the bridge assumes:

  • --input-device "CABLE-B Output"
  • --output-device "CABLE-A Input"

If your device names differ, list them first:

cd backend
uv run python -m src.lib.audio.remote_audio_server --list-devices

Then launch the server with explicit device names if needed:

cd backend
uv run python -m src.lib.audio.remote_audio_server --input-device "CABLE-B Output (VB-Audio Virtual Cable B)" --output-device "CABLE-A Input (VB-Audio Virtual Cable A)"

Also configure the game machine the same way as the local setup:

  • In VRChat, set Microphone to CABLE-A Output
  • In Windows, set the default playback device to CABLE-B Input

On the OpenNeuro Machine

Point both remote components at the game machine’s bridge:

  • RemoteAudioIn
    • server_url: ws://<game-machine-ip>:8765/ws/audio
    • sample_rate: 48000
    • channels: 2
    • frame_ms: 20
  • RemoteAudioOut
    • server_url: ws://<game-machine-ip>:8765/ws/audio
    • sample_rate: 48000
    • channels: 1
    • frame_ms: 20

For voice-to-voice VRChat flows:

  • AudioIn or RemoteAudioIn
  • VAD
  • ASR
  • LLM
  • TTS
  • AudioOut or RemoteAudioOut

If your downstream audio stack expects mono, that’s okay: AudioFrame conversion inside the pipeline will down-mix and resample automatically.

Introduction