VoiceCompose App Icon VoiceCompose
Menu-bar Voice Assistant for macOS

Speak a sentence, get a message ready to send.

Press Option+Space to record. VoiceCompose transcribes in real-time, polishes your text with AI, and inserts it straight into your focused app with Enter.

Coming Soon to the Mac App Store
Explore Features ↓

Coming Soon · Requires macOS 14+ & Apple Silicon

VoiceCompose Preferences
VoiceCompose Settings UI - Light Mode

Get into a Flow State in 4 Simple Steps

Fits naturally into your macOS workflow. No app switching, no manual copy-pasting.

01

1. Trigger Hotkey

Press ⌥ Space anywhere. A sleek floating panel appears instantly to stream real-time transcription.

⌥ Space
02

2. Speak Naturally

Talk freely in your natural voice without worrying about grammar, filler words, or punctuation.

Natural Voice Input
03

3. AI Polishing

Context-aware AI refines your thoughts. Hit Tab to switch between Clean Up, Chat, and Professional modes.

⇥ Tab
04

4. Insert & Send

Press Enter to insert the polished text into the active field and send it automatically.

↵ Enter

Three Tailored Modes for Every Scenario

Press Tab to switch styles on the fly — perfectly matched for casual chats, work updates, or notes.

整理

Clean Up Mode

Fixes typos, filler words, and punctuation while keeping your original intent. Ideal for quick notes.

聊天

Chat Mode

Transforms speech into natural, conversational messages perfect for Slack, WeChat, and daily messaging.

正式

Professional Mode

Refines input into clear, structured, and formal phrasing for emails, docs, and executive updates.

Engineered with Apple-Grade Precision

Every detail crafted specifically for Mac users who value speed and privacy.

Global Hotkey & Floating Panel

⌥ Space summons VoiceCompose instantly. Floating panel streams status and stays out of your way.

Dual-Engine Real-Time ASR

Powered by native Apple Speech Recognition for speed, or local Whisper (whisper.cpp) for fully offline transcription.

Three Polishing Modes

Tab between Clean Up, Chat, and Professional modes seamlessly to suit your destination context.

Smart Text Insertion & Restore

Inserts text directly into the active field via Accessibility API with clipboard fallback, automatically restoring original clipboard data.

Privacy-First Architecture

No audio files saved to disk. History stored in local JSON with 1-click wipe. No VoiceCompose servers.

Dual LLM Backend

Connect your custom OpenAI-compatible API (DeepSeek, GPT) or run completely offline using local Ollama.

Keyboard-Driven Efficiency

Stay in the zone. Control everything without taking your hands off the keyboard.

⌥ Space
Start / Stop recording
↵ Enter
Insert text & send
esc
Cancel operation
⇥ Tab
Switch polishing mode
R
Re-record voice input
Privacy First

Privacy Built-In, Not Added On

Your voice and text data remain fully under your control.

  • Zero Audio Retained: Audio is processed in memory for live transcription and never written to disk.
  • Full Offline Capability: Local whisper.cpp and Ollama support enable complete offline operation without internet.
  • No VoiceCompose Servers: Requests are sent directly to your configured LLM API with no telemetry middleman.
  • Transparent Permissions: Strictly requests only Microphone, Speech Recognition, and Accessibility.
Read Complete Privacy Policy →

Ready to Upgrade Your Daily Workflow?

VoiceCompose is currently under review for the Mac App Store.

Coming Soon to the Mac App Store
Get Support & Usage Help

System Requirements & Permissions

System

Operating System: macOS 14.0 (Sonoma) or later

Architecture

Architecture: Optimized for Apple Silicon (M1/M2/M3/M4), fully compatible with Intel processors

Required Permissions:

  • 🎤 Microphone: Microphone: Captures audio during recording
  • 🗣 Speech Recognition: Speech Recognition: Native Apple Speech transcription
  • ♿ Accessibility: Accessibility: Inserts text into the active text field and triggers Enter key send