Audio · Web

OSC ↔ VoiceMeeter bridge: a mixing console on your phone

A Python bridge turns any phone or tablet into a mixing console for VoiceMeeter Potato and a control panel for the PC: per-app audio, monitors, windows and more.

Year: 2026
Status: In daily use
Role: Design & development · personal project
Repository: Repository coming soon

01 · Problem

Mixing in VoiceMeeter Potato (5 hardware inputs, 3 virtual inputs and 8 buses) with a mouse is slow, and Windows does not let you send each application to a different virtual input without digging through menus. I wanted to control everything from my phone, in real time.

02 · Approach

I wrote a bidirectional OSC/UDP bridge in Python between Open Stage Control and VoiceMeeter. The UI (9 tabs) is never edited by hand: build scripts generate it, and some parts are generated from the detected hardware. Observer threads push the real state back to the screen.

03 · Result

Full mixing with live stereo VU meters, per-app volume and routing (it even migrates already-open audio sessions in about 2 s), a dashboard with CPU, RAM and temperature, monitor control over DDC/CI, mouse, keyboard, window management and PotPlayer remote. It starts automatically from the system tray.

04 · Stack

Python
python-osc
voicemeeterlib
pycaw / COM
Open Stage Control
DDC/CI
C# .NET 9 (diagnostics)

How it fits together

Phone / tablet (Open Stage Control, browser)
        │  OSC over UDP
        ▼
osc_bridge.py ──► VoiceMeeter Potato (mixing, buses, VU meters)
        │     ──► Windows audio sessions (per-app volume and routing)
        │     ──► Monitors (DDC/CI), mouse and keyboard (SendInput), windows
        ▲
   observer threads push the real state back to the UI

Decisions that mattered

The UI is compiled, not hand-drawn. Nine build_*.py scripts generate the Open Stage Control session in order, so design changes are reproducible and never lost between edits.
The monitors UI comes from the hardware. At build time the bridge probes what each monitor supports (brightness, contrast, color temperature, inputs). If a monitor supports picture modes but does not declare them, it discovers them by writing and reading back, and caches the result.
Audio comes first. Enumerating audio sessions faster than 1 Hz competed with the audio engine and caused dropouts, so the scan runs at 1 Hz.
Real per-app routing. Each app’s MUS / PEL / VOZ buttons (music, movies, voice) send it to the matching VoiceMeeter virtual input and reflect the actual routing when the tab opens.

Tabs

VoiceMeeter · Per-app volume · Dashboard · Monitors · Mouse · Keyboard (Latin American QWERTY with labels that react to Shift, Caps and AltGr) · Windows (cards, list or per-program views, with thumbnails) · PotPlayer · Smartbox.