An AI-powered, system-wide voice typing desktop application for Windows. Works instantly in any app.
Fluence is a lightweight, privacy-focused Windows application that brings human-level voice transcription to your desktop. Using the Groq Whisper API (whisper-large-v3) for near-instant, highly accurate transcription, Fluence sits quietly in your system tray and activates via global system hotkeys. When you speak, a floating visualizer overlay tracks your voice, and the transcribed text is injected directly at your cursor in whatever application you are using.
| 1. General Preferences | 2. AI Providers | 3. Custom Dictionary |
|---|---|---|
 Hotkey & Recording preferences |
 Groq / OpenAI STT and LLM configuration |
 Automatic text-corrections table |
- Completely Offline Mode: Supports local offline transcription using the SenseVoice-Small model powered by
sherpa-onnx, ensuring total privacy and internet-free voice typing. - On-Demand Model Manager: Built-in download manager in the settings dashboard to seamlessly download, monitor, or delete local transcription model files.
- Global System Shortcuts: Start and stop recording from any app with configurable hotkeys (default
Ctrl+Shift+Spacefor Voice Typing, andCtrl+Shift+Spacefor AI Agent Mode is configurable). - Flexible Recording Modes: Supports both Push-to-Toggle (press once to start, press again to stop) and Hold-to-Record (hold key down, speak, release to transcribe) recording styles.
- AI Polish Assistant: Choose from multiple language model polish styles to automatically clean up filler words, rewrite text in a professional business tone, convert spoken thoughts into bullet points, or translate dictation into clear English before pasting.
- Smart Highlight Grabber: In Agent Mode, the app can automatically grab currently selected text from your active application to use as context for edits, rewrites, or replies without changing your system clipboard.
- Copy to Clipboard Action: Command the AI agent to copy its output directly to your system clipboard instead of pasting it, complete with a clean visual success verification.
- Single-Instance Support: Built-in architecture that prevents multiple instances of Fluence from running. Launching a second instance automatically focuses and restores the active settings dashboard.
- Premium Fluent Design: An elegant, modern settings dashboard featuring branded purple and cyan ambient glows, custom typography, and a glassmorphic waveform visualizer matching Windows 11.
- Silent & Fast Injection: Injects transcribed text directly at your cursor via the native Windows
SendInputkeyboard API, preventing focus stealing and clipboard contamination. - Secure API Storage: Integrates with the Windows Credential Manager to encrypt and safely store your Groq API keys locally.
- SQLite History Database: Keeps a searchable local log of your past dictations for quick reuse or search.
- Word Corrections Dictionary: A customizable dictionary that automatically corrects specific phrases or misheard words (e.g., auto-capitalize names or project terms).
- Windows Auto-Start: Configure the application to run automatically on Windows boot, operating seamlessly in the background.
The Dictation Bottleneck: Built-in Windows dictation (Win+H) is often slow, struggles with accents, handles punctuation rigidly, and sends telemetry data to Microsoft servers. Other streaming voice tools transcribe word-by-word, missing the broader sentence context and frequently making homophone errors.
The Fluence Approach: Fluence captures your audio buffer, applies peak volume normalization to ensure quiet speech is clearly heard, encodes it into a highly optimized 64kbps MP3 stream to minimize network payload, and sends it directly to Groq's high-speed Whisper servers. It transcribes full sentences in under a second with intelligent punctuation and deep context awareness, inserting it instantly at your cursor.
| Feature | Fluence Windows | Windows Built-In Speech |
|---|---|---|
| Accuracy | 95% - 98% (Human level) | 80% - 88% |
| Punctuation | Automatic, Contextual & Intelligent | Rigid / Manual commands |
| Privacy | Open Source (Zero Telemetry) | Closed Source (Telemetry sent) |
| Integrates Everywhere | Yes (Direct text input injection) | Limited to specific input boxes |
| Response Latency | Sub-second (~700ms) | Variable |
| Custom Dictionary | Yes (Auto-corrections) | Limited |
Fluence is built for users who prioritize privacy and local data security:
- 100% Offline Dictation Option: When using the local offline provider, your audio never leaves your system. Transcription runs completely locally on your hardware.
- Zero Telemetry or Phone-Home Code: No tracking code, analytics scripts, background logging, or usage telemetry. The code is fully open source and verifiable.
- Direct HTTPS Transmission: Audio goes directly from your local machine to the API endpoints (e.g.,
https://api.groq.com). No intermediate servers or proxies. - Windows Credential Manager Storage: Your secret API keys are stored using the Windows OS native secure credential store (
wincred), encrypted via Windows DPAPI. - Ephemeral Audio Storage: Audio recordings are processed on the fly in system memory, converted to MP3, and wiped immediately once transcription is complete.
- Download the latest installer (
.exeor.msi) from the Releases tab. - Run the installer and open Fluence.
- The onboarding setup wizard will guide you to configure your Groq API Key.
- Press
Ctrl+Shift+Spacein any app, or set your favorite custom shortcut, speak, and press it again to watch your voice convert to text instantly!
To run or build the application from source, you need Rust and Node.js installed on your system.
- Clone the repository.
- Install the dev dependencies:
npm install
- Run the development environment:
npm run dev
- Build the production installers (Wix MSI and NSIS EXE):
The installers will be generated inside
npm run build
src-tauri/target/release/bundle/.
Fluence is licensed under the GNU Affero General Public License v3.0 only (AGPL-3.0-only).
See the LICENSE file for the complete license text.