🎧 Application Audio Capture (Windows Only)

N-API port of the WASAPI process-loopback sample to get a nodejs Buffer of an application's audio - or of the entire system's audio.

⚠️ Requirements: This package only runs on Windows 10 x64 and later. It uses native binaries and will warn on unsupported platforms.

🚀 Features

• Capture raw PCM audio from individual applications using WASAPI process-loopback.
• Capture raw PCM audio from the entire system (whatever's playing through the default output device) - no process ID required.
• Pipe real-time audio data into your JavaScript/TypeScript app.

https://github.com/user-attachments/assets/fc058596-6ea3-4ded-8065-aeb642c5a465

📦 Installation

Install using your favorite package manager.

npm install application-loopback
#OR
bun install application-loopback
#OR any package manager..

🧠 Usage

1. Start Capturing Audio from a Single Process

import loopback from "loopback-capture";

const capture = new loopback.LoopbackCapture();
const processId = capture.start(1234, /* includeProcessTree */ true, (chunk: Buffer) => {
  console.log("Application audio data:", chunk); // Buffer
});

process.on("SIGINT", () => {
  capture.stop();
  process.exit(0);
});

🧠 chunk is a raw PCM audio buffer. You can pipe it to a file, stream it, analyze it, etc.

2. Start Capturing System-Wide Audio (no PID needed)

Want everything the system is playing - not just one app? Use startSystemAudio instead. It skips process filtering entirely and captures straight from the default playback device.

import loopback from "loopback-capture";

const capture = new loopback.LoopbackCapture();
capture.startSystemAudio((chunk: Buffer) => {
  console.log("System audio data:", chunk); // Buffer
});

process.on("SIGINT", () => {
  capture.stop();
  process.exit(0);
});

🧠 Same raw PCM format as process capture (16-bit, stereo, 48kHz) - just sourced from the whole desktop instead of one process tree.

🪟 Why and how?

For my desktop application (Huginn) I needed to capture an application's audio selectively to share in a call just like discord does. I later added full system audio capture too, for cases where selecting one process isn't the point - like recording the whole desktop for a meeting.

The two modes use different Windows APIs under the hood:

• Per-process capture uses AUDIOCLIENT_ACTIVATION_TYPE_PROCESS_LOOPBACK (Windows 10 2004+), activated asynchronously via ActivateAudioInterfaceAsync, and can optionally include a process's child tree.
• System-wide capture uses the classic WASAPI loopback-recording technique - grabbing the current default playback device via IMMDeviceEnumerator and activating an IAudioClient on it directly with AUDCLNT_STREAMFLAGS_LOOPBACK - the same approach tools like OBS use for desktop audio.

Most of the C++ source is simply a stripped out version from a sample in microsoft's classic samples repo https://github.com/microsoft/Windows-classic-samples

🧪 Example Use Cases

• Build a real-time audio visualizer for specific apps.
• Record browser or game audio selectively.
• Stream audio from only one process instead of the whole system.
• Record full desktop/system audio for meeting or screen recordings.