Voco

Voice input that understands you.
Speak naturally, get clean text pasted anywhere on your Mac.

---

Voco is a lightweight macOS menu bar app that turns your voice into polished, ready-to-use text. Press a hotkey to start recording, press again to stop — Voco transcribes your speech, cleans it up with an LLM, and pastes the result directly into whatever app you're using.

   

Features

Two Modes

• Transcribe (Fn) — Records your voice, cleans up filler words, fixes grammar, and pastes polished text
• Translate (Shift+Fn) — Everything above, plus translates from one language to another (e.g. Chinese to English)

Both hotkeys are fully customizable — supports Fn, F1–F12, and any modifier combo.

Smart Processing

• Context-aware tones — Detects the active app (Slack, Mail, Xcode, etc.) and adjusts tone automatically. Choose from formal, casual, technical, neutral, concise, or friendly. Add custom tones for any app by bundle ID.
• Voice commands — Say "new line", "period", "question mark", "bullet list", or "第一/第二/第三" and they become real formatting
• Self-correction handling — When you rephrase mid-sentence, Voco keeps only the final version
• Filler word removal — Strips "um", "uh", "like", "you know", "嗯", "那个", etc.
• Personal dictionary — Add custom vocabulary with pronunciation hints and categories (Name, Technical, Brand, Acronym) so names and jargon are always spelled correctly
• Custom system prompts — Edit the transcribe and translate prompts in your text editor with live reload. Use placeholders like {{app_name}}, {{tone}}, and {{dictionary}} for dynamic context.

Designed for macOS

• Lives in the menu bar — always accessible from the top of your screen
• Floating overlay shows recording status with real-time audio level
• Global hotkeys work from any app
• Direct paste via keyboard simulation — text appears right where your cursor is
• Retry last recording if the result wasn't right
• Transcription history with timestamps
• Launch at login support
• Audio feedback sounds for recording start/stop/done/error

Bring Your Own API

Voco uses cloud APIs with OpenAI-compatible endpoints. STT and LLM are configured independently — different providers, endpoints, models, and API keys for each:

Component	Default	Alternatives
Speech-to-Text	Qwen ASR (DashScope)	Any Whisper-compatible API (OpenAI, Groq, etc.)
LLM	GPT-4o-mini (OpenAI)	Cerebras, Groq, or any OpenAI-compatible endpoint

Installation

Requirements

• macOS 15 (Sequoia) or later
• Microphone permission
• Accessibility permission (for pasting text into other apps)
• API keys for your chosen STT and LLM providers

Homebrew

brew tap Stanleytowne/voco
brew install --cask voco

Download

Grab the latest .dmg from Releases, open it, and drag Voco to your Applications folder.

Note: Since the app is not notarized, macOS may block it on first launch. Right-click the app and select Open to bypass Gatekeeper.

Build from Source

# Clone the repository
git clone https://github.com/Stanleytowne/Voco.git
cd Voco

# Build (requires Xcode, not just Command Line Tools)
cd Voco
swift build

# Deploy to app bundle
cp .build/debug/Voco .build/Voco.app/Contents/MacOS/

# Launch
open .build/Voco.app

Note: The executable must run inside the .app bundle. Running the binary directly will exit immediately because the app requires Info.plist (with LSUIElement for menu bar mode).

Getting Started

On first launch, Voco walks you through a setup wizard:

1. Grant microphone access — needed to record your voice
2. Grant accessibility access — needed to paste text into other apps
3. Enter STT API key — for your speech-to-text provider
4. Enter LLM API key — for your language model provider
5. Review hotkeys — defaults are Fn (Transcribe) and Shift+Fn (Translate)

Usage

1. Focus the app where you want to type
2. Press the hotkey (Fn to transcribe, Shift+Fn to translate)
3. Speak naturally — a floating overlay shows you're recording
4. Press the hotkey again (or click the checkmark on the overlay) to stop
5. Your text is automatically pasted at the cursor position

Press the X button on the overlay or Esc to cancel at any time. If the result wasn't right, use Retry Last Recording from the menu bar to reprocess the same audio.

Settings

Access Settings from the menu bar icon (Cmd+,).

General

Hotkeys, overlay toggle, audio feedback sounds, always-copy-to-clipboard option, launch at login, and permission status.

Speech (STT)

Configure your speech-to-text provider — base URL, model name, API key, and input language. Defaults to DashScope Qwen ASR.

AI (LLM)

Configure your language model provider — base URL, model name, and API key. Set translation languages (source and target). Edit the system prompts for both transcribe and translate modes — prompts open in your text editor and auto-reload on save. Includes a "Test Processing" button to verify your LLM connection.

Tones

Per-app tone profiles that automatically adjust the writing style based on which app is focused. Built-in profiles for common apps (Slack, Mail, Xcode, etc.) can be customized, and you can add new ones by app name and bundle ID.

Dictionary

Custom vocabulary with pronunciation hints. Add names, technical terms, brands, and acronyms with optional "sounds like" hints so the LLM always spells them correctly. Searchable and categorized.

History

All transcriptions are saved with timestamps. View them from the menu bar via History.

How It Works

Hotkey pressed → Start recording
    ↓
Hotkey pressed again → Stop recording
    ↓
Audio sent to STT API → Raw transcription
    ↓
Dictionary lookup → Correct custom vocabulary
    ↓
Detect active app → Select tone profile
    ↓
LLM processes text → Clean, formatted output
    ↓
Text pasted at cursor position
    ↓
Saved to history

License

MIT