Skip to content

anesask/voxring

Repository files navigation

VoxRing

Hold. Speak. Release. Your voice, anywhere.

VoxRing is a voice-to-everything plugin for Logitech MX Master 4 and MX Creative Console. It turns a single mouse button into a dictation engine, an AI writing assistant, a voice commander, and a conversational AI: all offline by default.

Built for the Logitech DevStudio Challenge 2026, Actions SDK category. By Anes Mulalic.


What it does

Hold the side button, speak, release. VoxRing transcribes your voice and routes the result to wherever you need it: clipboard, the focused window, email, Slack, Discord, Teams, or calendar. AI reformats the transcript for each destination automatically. No keyboard required.

36 actions across 8 groups:

  • Voice: Dictate, Push to Talk, Voice Assistant, Voice Note
  • Send: AI-formatted send, transcript history folder
  • Quick Send: one-gesture voice-to-destination for 7 targets
  • Text Tools: AI toggle, Noise Gate, Filler Cleaner, Case Transform, Translate, Append, Edit, Regenerate
  • Controls: Read Back, Replay Last, Voice Notes folder, Cycle Destination, Mute Mic, Cancel, Open Voice Notes
  • Settings: Language Toggle, full settings panel
  • Meters: live audio level dial, rolling waveform, word count
  • Utilities: Guitar Tuner easter egg

Three offline speech engines: Whisper (accurate; Tiny / Base / Small, downloadable in-app), Vosk (fast, ~40 MB), and Windows Speech (instant result on release, uses your installed Windows language pack). A built-in model catalogue lets you browse, download, select, and delete models from the settings page.

Destinations: Clipboard, Type Out (types into the focused window), Email, Slack, Discord, Teams, Calendar, plus unlimited custom Slack/Discord webhook channels with per-channel bot name and avatar. WhatsApp, Telegram, and Notion are coming soon.

AI formatting: Claude and OpenAI are live; Gemini, DeepSeek, and Perplexity are coming soon. All behind a single interface.


Installation

Download VoxRing-1.1.0.lplug4 and double-click to install via Logi Options+. Requires Logi Options+ 6.0 or later. Windows only.

On first launch, VoxRing downloads a speech recognition model and installs Piper TTS in the background. Recording is available immediately using the built-in Vosk model; Whisper loads once its download completes. Additional Whisper and Vosk models can be downloaded on demand from the Speech Models section in settings.


Setup

Open the VoxRing Settings action in Logi Options+ to configure:

  • Microphone: select your preferred input device
  • Speech model: pick an engine and model in the Speech Models catalogue (Whisper Tiny/Base/Small, Vosk, or Windows Speech); download and switch in place
  • Language: Auto-detect, English, or German
  • AI provider: paste your API key for Claude or OpenAI
  • Destinations: Slack and Discord webhook URLs, Teams webhook URL, plus custom labelled webhook channels

All API keys and webhook URLs are encrypted at rest using Windows DPAPI.


How it works

Voice is captured via NAudio at 16 kHz. Speech recognition runs fully offline using Vosk, Whisper.net, or the Windows SAPI recognizer: no audio leaves your machine unless you opt into cloud AI formatting. When AI formatting is enabled, the transcript is sent to your chosen provider with a destination-specific prompt (email tone for email, Slack style for Slack, etc.). Text-to-speech for Voice Assistant mode uses Piper neural voices offline.

See the repository wiki for architecture details, extension points, and how to add new destinations or AI providers.


Building from source

Prerequisites: .NET 8 SDK, Logi Options+ installed.

git clone https://github.com/anesask/voxring.git
cd voxring
dotnet build src/VoxRingPlugin.csproj

The post-build step deploys the plugin to Logi Plugin Service automatically. Restart Logi Options+ to load the updated plugin.

For VS Code users, the .vscode/ folder includes a pre-configured build task and a "Attach to Logi Plugin Service" debug launch configuration.


Third-party components

VoxRing is built on the following open-source libraries. Full license texts are in NOTICES.md.

Component License Purpose
NAudio MIT Microphone capture
Vosk Apache 2.0 Offline speech recognition (fast engine)
Vosk model (en-us-small) Apache 2.0 English speech model
Whisper.net MIT Offline speech recognition (accurate engine)
whisper.cpp MIT Native Whisper inference
OpenAI Whisper model weights MIT Speech model weights
System.Speech MIT Windows SAPI recognition (instant engine)
Silero VAD MIT Voice activity detection (auto-stop)
ONNX Runtime MIT Silero VAD inference
Piper TTS MIT Offline neural text-to-speech
Piper voice models CC0 1.0 English (Amy) and German (Thorsten) voices
Microsoft.Data.Sqlite MIT Local telemetry storage
Logi Actions SDK Logitech SDK License Plugin API

License

MIT License. Copyright (c) 2026 Anes Mulalic. See LICENSE for details.

Third-party components are subject to their own licenses as listed in NOTICES.md.

About

Voice-to-everything plugin for Logitech MX Master 4 and MX Creative Console. Dictate, send to 9 destinations, AI formatting, offline speech recognition.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors