Mike

AI-powered voice dictation for Windows — speak anywhere, instantly.

What is Mike?

Mike is a global voice dictation engine that runs silently in your Windows system tray and transcribes your speech into any application — email, browser, code editor, Slack, anywhere. No clicks, no switching focus. Just press a hotkey and talk.

It uses Groq's Whisper API for near-instant transcription (typically under 500ms) and optionally polishes your speech through an AI language model to produce clean, professional text.

Mike runs at startup, lives in the tray, and stays out of your way.

Features

Feature	Details
🎙️ Push-to-Talk	Hold `Ctrl+Shift`, speak, release — text appears at cursor
🔁 Continuous Mode	Toggle `Ctrl+Shift+Space` — mic stays live, transcribes every 5s
🧠 AI Polishing	Three modes: Raw, Semi-formal, Polished — cleans up your speech
💬 Symbol Expansion	Say "degree symbol" → `°`, "plus minus" → `±`, "copyright symbol" → `©`
🖥️ Floating HUD	Minimal pill above taskbar shows recording state and waveform
📊 Dashboard	Full history, session stats, settings — one click from tray
🚀 Startup	Registers itself to launch with Windows — always ready
🪟 Silent	No terminal, no console — runs invisibly in the background

Hotkeys

Action	Hotkey
Push-to-Talk (hold to record)	Ctrl + Shift
Toggle Continuous Mode	Ctrl + Shift + Space
Open Dashboard	Click tray icon → Dashboard
Cycle dictation mode	Click mode badge on HUD

HUD States

The floating pill above your taskbar tells you what Mike is doing:

State	Appearance
Silent	Tiny dark pill with mic dot — app is alive and listening
Hover	Expands to show hotkey hint and controls
Recording	Waveform animation — actively capturing audio
Continuous	Green border, LIVE badge, pulsing dot
Processing	Spinner — transcribing and injecting text

Dictation Modes

Mode	What it does
RAW	Transcribes exactly what you say, minimal processing
SF (Semi-formal)	Cleans grammar, removes filler words, professional tone
POL (Polished)	Full AI rewrite — structured, publication-ready prose

Switch modes by clicking the RAW / SF / POL badge on the HUD, or via the Dashboard Settings tab.

Symbol Expansion

Say any of these phrases and Mike will type the actual symbol:

Say	Get	Say	Get
"plus minus symbol"	±	"copyright symbol"	©
"degree symbol"	°	"trademark symbol"	™
"right arrow"	→	"check mark"	✓
"infinity symbol"	∞	"not equal"	≠
"square root"	√	"approximately"	≈
"euro symbol"	€	"bullet point"	•
"em dash"	—	"dot dot dot"	…

Requirements

OS: Windows 10 or 11
Python: 3.11+ (for building from source)
Microphone: Any input device supported by Windows
API Key: Groq API key (free tier available)

How to get a Groq API Key

Go to the Groq Console and create a free account.
Click on API Keys in the left sidebar.
Click the Create API Key button.
Name your key (e.g., "Mike Dictation") and copy the generated key (it starts with gsk_...).
Open Mike's Dashboard → Settings, paste your key, and click Save.

Installation

Option A — Build from source (recommended)

1. Clone the repo

git clone https://github.com/yourusername/mike.git
cd mike

2. Set up your API key

cp config.example.json config.json
# Edit config.json and paste your Groq API key

Get a free Groq API key at console.groq.com

3. Run the build & install script

powershell -ExecutionPolicy Bypass -File build_and_install.ps1

This script will:

Install Python dependencies in a virtualenv
Generate crisp icons from SVG sources
Build Mike.exe via PyInstaller (~91MB, fully self-contained)
Install to %LOCALAPPDATA%\Programs\Mike\
Create Desktop and Start Menu shortcuts
Register Mike to launch at Windows startup
Start Mike immediately

Total time: ~3 minutes

Option B — Pre-built binary

Download Mike.exe from Releases and run it. On first launch it will ask for your Groq API key via the Dashboard settings.

First Run

After installation:

Mike starts automatically and appears as a small dark pill near the center of your taskbar
Click the tray icon (bottom-right) to open the Dashboard
Go to Settings → paste your Groq API key → Save
Press Ctrl+Shift and start talking

How It Works

You speak
    ↓
Microphone (sounddevice, 16kHz mono)
    ↓
Voice Activity Detection (RMS threshold filter)
    ↓
Groq Whisper API (transcription, <500ms)
    ↓
Text Filter (filler removal, symbol expansion, dedup)
    ↓
AI Polish (optional — Groq LLaMA via chat completions)
    ↓
pyautogui.typewrite() → text appears at cursor

Architecture

File	Role
`main.py`	Entry point, startup flags, single-instance lock, orchestration
`engine.py`	State machine — IDLE / PTT / CONTINUOUS / PROCESSING
`hotkeys.py`	Global hotkey listener (pynput), timer-deferred PTT
`audio.py`	Mic capture, VAD, 16kHz WAV encoding
`transcription.py`	Groq Whisper API client
`ai.py`	Groq LLaMA polishing (semi-formal / polished modes)
`filters.py`	Text cleaning, symbol expansion, filler removal
`injection.py`	Keyboard text injection via pyautogui
`hud.py`	Floating Tkinter HUD pill — 4 animated states
`dashboard.py`	PyQt6 dashboard — history, stats, settings
`tray.py`	pystray system tray icon and menu
`db.py`	SQLite history database
`startup.py`	Windows registry auto-start

Configuration

config.json (created at %LOCALAPPDATA%\Mike\config.json):

{
  "groq_api_key": "gsk_...",
  "default_mode": "semi_formal",
  "hud_opacity": 0.85,
  "hud_x": null,
  "hud_y": null,
  "continuous_chunk_seconds": 5,
  "transcription_language": "en",
  "inject_method": "clipboard"
}

Key	Values	Default
`groq_api_key`	Your Groq API key	—
`default_mode`	`raw` / `semi_formal` / `polished`	`semi_formal`
`hud_opacity`	`0.1`-`1.0`	`0.85`
`continuous_chunk_seconds`	`3`–`10`	`5`
`transcription_language`	`en` / `hi` / `ur` / any Whisper lang code	`en`
`inject_method`	`clipboard` / `type`	`clipboard`

Uninstall

powershell -ExecutionPolicy Bypass -File uninstall.ps1

This removes the exe, shortcuts, and registry startup entry. Your config and history are preserved.

Troubleshooting

Mike doesn't start

Check %LOCALAPPDATA%\Mike\mike.log for errors
Ensure your Groq API key is valid at console.groq.com

Hotkeys not triggering

Make sure Mike is running (check system tray)
Ctrl+Shift may be grabbed by Windows language switcher if you have multiple input languages — try removing extra languages from Windows settings

Audio rejected / nothing transcribed

Speak for at least 0.3 seconds
Check your default microphone in Windows Sound settings

Text appears in wrong place

Click your target text field first, then press the hotkey

Building

Dependencies are listed in requirements.txt. Key ones:

groq
pynput
PyQt6
pystray
sounddevice
scipy
numpy
pyautogui
cairosvg
pillow
pyinstaller

License

MIT — see LICENSE

Built with ❤️ using Groq · Whisper · Python

Made by h55n

Name		Name	Last commit message	Last commit date
assets		assets
docs		docs
scripts		scripts
src		src
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
config.example.json		config.example.json
mike initiation sound.mp3		mike initiation sound.mp3
mike-design.md		mike-design.md
mike.spec		mike.spec
requirements.txt		requirements.txt

Navigation Menu

Repository files navigation

Mike

What is Mike?

Features

Hotkeys

HUD States

Dictation Modes

Symbol Expansion

Requirements

How to get a Groq API Key

Installation

Option A — Build from source (recommended)

Option B — Pre-built binary

First Run

How It Works

Architecture

Configuration

Uninstall

Troubleshooting

Building

License

About

Releases

Packages

Contributors

Languages

Latest commit

History

Folders and files

Repository files navigation

Mike

What is Mike?

Features

Hotkeys

HUD States

Dictation Modes

Symbol Expansion

Requirements

How to get a Groq API Key

Installation

Option A — Build from source (recommended)

Option B — Pre-built binary

First Run

How It Works

Architecture

Configuration

Uninstall

Troubleshooting

Building

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages