Cable

한국어 README

A real-time audio routing desktop app for Windows.Build a node graph that connects physical and virtual audio devices, and create virtual audio endpoints on demand through a custom kernel driver.

Warning Cable is in early development. The virtual audio device feature requires Windows test signing to be enabled.

---

Project Summary

Cable lets you wire any audio source to any audio sink on Windows from a visual node graph. Built-in nodes cover physical input/output devices (via cpal), virtual devices created by the bundled CableAudio kernel driver, app loopback capture (WASAPI), VST3 plugin hosting, and a small library of utility processors (Mixer, Gain, Channel Split/Merge, Delay, Compressor, Reverb, Echo, Spectrum Analyzer, Waveform Monitor).

The app is built on Tauri v2: a React + React Flow frontend drives a Rust audio runtime that processes the graph in topological order on a sub-millisecond spin-loop thread. A type system on the frontend statically validates each connection (channels / sample rate / bit depth) before the graph is pushed to the runtime, so only fully-validated graphs ever reach the audio engine.

---

Getting Started

Install a release

1. Download the latest Cable_<version>_x64-setup.exe from the Releases page.
2. Run the installer.
3. (Optional, for virtual audio devices) Enable Windows test signing — see Notes.
4. Launch Cable from the Start Menu.

The first launch creates two placeholder nodes (Audio Input Device → Audio Output Device). Right-click the canvas to add more nodes; drag from one handle to another to connect.

System requirements

• Windows 10 (1809+) or Windows 11
• x64 CPU
• For virtual audio devices: test-signing mode enabled and the CableAudio driver installed

---

Features

• Visual node graph — React Flow canvas with drag-to-connect, right-click context menus, and per-edge type badges
• Physical I/O — Enumerate and route any WASAPI / DirectSound / WDM device through cpal
• Virtual audio devices — Create render and capture endpoints on the fly via the CableAudio kernel driver; rename them with UAC elevation
• App loopback capture — Capture audio from a specific running process (per-app capture, Windows 10 2004+)
• VST3 host — Load and run third-party VST3 plugins inside the graph
• Built-in processors — Mixer, Gain, Channel Split / Merge (2/4/6/8 channels), Delay, Compressor, Reverb, Echo
• Visualizers — Spectrum Analyzer (FFT) and Waveform Monitor (oscilloscope-style)
• Static type checking — Every edge is validated against the format expected by its sink; mismatched edges are highlighted and the graph is held back from the runtime until valid
• Save / load — Persist a graph as JSON and drag-and-drop to reload
• Sub-millisecond runtime — Topological-order spin-loop processing thread for low-jitter audio routing

---

Notes

• Test signing for virtual devices. The CableAudio kernel driver is currently signed with a self-generated test certificate. To install it you must enable test mode:

  bcdedit /set testsigning on

  followed by a reboot. A production-signed driver is on the roadmap.
• Buffer size. The runtime defaults to a 512-frame buffer at the device's native sample rate. Smaller buffers reduce latency but increase the chance of underruns on heavily loaded graphs.
• Virtual device naming. Renaming a virtual endpoint writes to PKEY_Device_FriendlyName, which requires administrator privileges. Cable re-launches itself with elevation only for the rename action; the main window does not need admin rights.
• Graph validity gate. If any node fails validation (e.g. a stereo source feeding a mono sink), the runtime keeps running the last fully-valid graph. Fix the offending edge or node and the new graph is pushed automatically.
• Licensing. App code (src/, crates/) is GPL-3.0; the kernel driver (driver/) is MS-PL, derived from the Microsoft WDK samples and VirtualDrivers/Virtual-Audio-Driver. See LICENSE.

---

Build Instructions

Prerequisites

• Rust (stable, with x86_64-pc-windows-msvc)
• Node.js ≥ 20
• pnpm ≥ 10
• Visual Studio Build Tools with the Desktop development with C++ workload
• Windows Driver Kit (WDK) — only needed if you build the driver

Common workflows

# Install JS deps
pnpm install

# Run the app in dev mode (frontend HMR + Tauri)
pnpm tauri dev

# Frontend-only dev server (no Rust)
pnpm dev

# Production app build
pnpm tauri build

# One-shot full build (driver + frontend + Tauri app)
.\scripts\build.ps1

# Driver only (requires WDK)
.\scripts\build.ps1 -Target Driver

# App only (skip driver)
.\scripts\build.ps1 -Target App

Quality gates

pnpm lint            # oxlint
pnpm fmt:check       # oxfmt --check
pnpm test            # Vitest (frontend)
cargo test --workspace   # Rust unit + ABI tests
cargo fmt            # rustfmt

VM-based driver tests

Driver integration tests run inside a VMware guest. Prerequisites: VMware Workstation + vmrun on PATH, driver build artifacts under driver/x64/Debug/package/, and a .env with VM_PASSWORD=....

.\.vm\setup.ps1                      # one-time VM provisioning
.\.vm\test.ps1                       # run all Pester suites
.\.vm\test.ps1 -TestFilter "*IOCTL*" # filter by Pester FullName
.\.vm\exec.ps1 "Get-PnpDevice -Class MEDIA"

---

Roadmap

• Production-signed kernel driver (no more test-signing required)
• Per-edge sample rate conversion and bit-depth dithering
• MIDI-style modulation graph (control-rate edges) alongside audio edges
• More built-in processors: parametric EQ, limiter, noise gate, convolution reverb
• LV2 / CLAP plugin hosts in addition to VST3
• Multi-graph project files with per-graph enable/disable
• macOS port (Core Audio backend; virtual device support TBD)
• Real-time CPU and latency telemetry per node