Typeflux is a macOS menu bar application designed for seamless voice-to-text input, enabling users to dictate text directly into any application with a "hold to talk, release to insert" workflow. It targets power users and anyone seeking to significantly accelerate their typing speed and reduce context switching, offering a privacy-first, local-first approach to speech recognition and AI-powered text manipulation.

How It Works

Typeflux operates on a simple, intuitive hotkey-driven model. Users press and hold a designated hotkey (defaulting to Fn) to begin dictation, speak naturally, and release the key to have the transcribed text instantly injected at the cursor's current position in the active application. A double-press of the hotkey activates an "Ask Anything" mode, allowing voice commands for AI-driven tasks like Q&A, content rewriting, or complex workflow execution. This direct injection mechanism bypasses the need to switch between applications, preserving user flow and dramatically increasing input efficiency. The system prioritizes local inference, running models directly on the user's Mac for enhanced privacy and offline capability.

Quick Start & Requirements

• Installation: Download the latest release .dmg file and drag Typeflux.app to the Applications folder.
• Prerequisites: macOS 13 or later, Microphone permission, Accessibility permission. For cloud providers, API keys and endpoint URLs are required. Local inference models are downloaded automatically.
• Build from Source: Clone the repository (git clone https://github.com/mylxsw/typeflux), run scripts/setup_dev_cert.sh for persistent macOS permissions, then make run to build and launch.
• Links: Typeflux Website, Screenshots, Video Demo

Highlighted Details

• Direct Text Injection: Dictated text is inserted directly into the active application at the cursor, eliminating copy-pasting.
• "Ask Anything" AI Integration: Double-tap hotkey to engage AI for Q&A, text editing, and workflow automation.
• Local-First & Privacy: Supports on-device models for offline use and ensures no voice or text data leaves the user's machine.
• Multiple Speech Backends: Offers flexibility with cloud providers (Google, OpenAI, Alibaba, etc.) and various local models (WhisperKit, SenseVoice, FunASR).
• Custom Personas: Define and switch between named instruction sets for different contexts (e.g., emails, code comments).
• Streaming Preview: Displays partial transcriptions in real-time while speaking.
• History & Replay: Saves all sessions locally for review, re-transcription, or export.

Maintenance & Community

Typeflux is an open-source project welcoming community contributions. Development updates and discussions can be followed via their Discord server and WeChat group.

Licensing & Compatibility

• License: AGPL-3.0. This copyleft license requires derivative works to be made available under the same license.
• Compatibility: Strictly macOS-only. The AGPL-3.0 license may impose restrictions on integration with closed-source commercial applications.

Limitations & Caveats

The application is exclusively for macOS. The AGPL-3.0 license necessitates that any modifications or integrations distributed must also be open-sourced under AGPL-3.0, which could be a significant consideration for commercial adoption or closed-source projects. Microphone and Accessibility permissions are mandatory for core functionality.