VoiceWin

Native Windows Speech-to-Text Transcription App
Press a key, speak, release — your words appear instantly.

Download • Quick Start • Features

---

Features

• Hold Right Alt to record, release to transcribe and paste
• Tap Right Alt to toggle recording (configurable)
• Hybrid mode — hold for quick recordings, tap to toggle for longer ones
• Real-time Streaming — see transcriptions appear as you speak (Deepgram)
• Status Overlay — animated waveform bars with REC/LIVE indicator, shows Recording/Listening feedback
• Voice Activity Detection — automatically trims silence, auto-stops streaming after extended silence
• AI Enhancement — polish transcriptions with LLM-powered rewriting
• 20+ Languages — supports English, Spanish, French, German, Japanese, Chinese, and more
• Sound Feedback — audio cues when recording starts/stops
• Tray Icon Status — visual indicator shows recording (red), processing (orange), or ready (green)
• Supports Groq Whisper API (fast, free tier available)
• Supports Deepgram nova-3 (high quality, non-streaming or streaming)
• System tray app — runs in background
• Auto-paste transcribed text to focused window
• Single portable EXE — no installation required

Requirements

• Windows 10/11 x64

Quick Start

1. Download VoiceWin.exe from Releases
2. Run it — no installation needed
3. Add your API key (Groq or Deepgram) in the settings window
4. Press Right Alt to record, release to transcribe

API Keys

Groq (Recommended — Fast & Free)

1. Go to console.groq.com
2. Create an account and get an API key
3. Paste in the "Groq API Key" field

Deepgram

1. Go to console.deepgram.com
2. Create an account and get an API key
3. Paste in the "Deepgram API Key" field

Usage

1. Press and hold Right Alt to start recording
2. Speak into your microphone
3. Release Right Alt to stop and transcribe
4. Text is automatically pasted into the focused text field

Status Overlay

A minimalist overlay appears during recording:

• Animated waveform — 5 bars that react to your microphone audio level
• REC indicator (red) — shown during batch recording mode
• LIVE indicator (green) — shown during streaming mode
• Recording/Listening — real-time feedback showing when speech is detected
• Configurable position — top or bottom of screen

The overlay automatically hides after transcription completes.

Voice Activity Detection (VAD)

Powered by Silero VAD for intelligent silence handling:

Batch Mode

• Automatically trims silence from the beginning, middle, and end of recordings
• Reduces API costs by sending only speech segments
• Shows "No speech detected" if recording contains only silence

Streaming Mode

• Auto-stops recording after configurable silence timeout (default: 60 seconds)
• Prevents overnight API charges if you forget to stop recording
• Configurable in settings

Setting	Default	Description
VAD Enabled	On	Toggle silence detection
Silence Timeout	60s	Auto-stop streaming after this much silence

Hotkey Modes

Mode	Description
Hold to Record	Hold the key while speaking, release to transcribe
Tap to Toggle	Tap once to start, tap again to stop
Hybrid	Hold for quick recordings (≥250ms), or tap to toggle for longer sessions

Transcription Providers

Non-Streaming (Groq, Deepgram)

Records audio, then transcribes after you stop. Fast and reliable.

Streaming (Deepgram)

Transcribes in real-time as you speak — text appears immediately. Best used with Toggle hotkey mode.

Note: There's a ~1 second delay while connecting to the Deepgram WebSocket. The sound plays once connected — wait for it before speaking.

Note: Some apps (terminals, code editors) may strip trailing whitespace between transcript chunks. If words run together, try non-streaming mode or adjust your app's settings.

Language Support

VoiceWin supports 20+ languages with auto-detection:

Language	Code	Language	Code
Auto-detect (Multi)	multi	Japanese	ja
English	en	Korean	ko
Spanish	es	Russian	ru
French	fr	Portuguese	pt
German	de	Italian	it
Chinese	zh	Dutch	nl
Hindi	hi	Polish	pl
Arabic	ar	Turkish	tr

Select your preferred language in settings, or use "Auto-detect" for multilingual transcription.

AI Enhancement

Enable AI Enhancement to polish transcriptions with Groq's LLM:

• Fixes grammar, spelling, and punctuation
• Removes filler words and stutters
• Improves sentence structure and clarity
• Preserves your original meaning and tone

Customize the enhancement prompt in settings.

Tray Icon Status

Color	Status
🟢 Green	Ready
🔴 Red	Recording
🟠 Orange	Processing

Settings

Settings are stored at:

%APPDATA%\VoiceWin\settings.json

Setting	Default	Description
Transcription Provider	Groq	Groq, Deepgram, or Deepgram Streaming
Hotkey Mode	Hold	Hold, Toggle, or Hybrid
Language	Auto-detect	20+ languages supported
Overlay Position	Bottom	Top or Bottom of screen
VAD Enabled	On	Silence detection and trimming
VAD Silence Timeout	60s	Auto-stop streaming after silence
AI Enhancement	Off	LLM-powered text cleanup
Sound Feedback	On	Audio cues for recording

Build from Source

Requires .NET 8.0 SDK.

# Development build
dotnet restore
dotnet build src/VoiceWin -c Release

# Self-contained single-file EXE
dotnet publish src/VoiceWin -c Release -r win-x64 --self-contained true -p:PublishSingleFile=true -p:IncludeNativeLibrariesForSelfExtract=true -o publish

Output: publish/VoiceWin.exe (~560MB due to bundled ONNX Runtime for VAD)

Note: The large file size is due to ONNX Runtime CUDA libraries bundled for Silero VAD. The app works without GPU acceleration.

License

MIT