SpeechForge Dictation

An AI-powered speech dictation desktop application that runs completely locally — no cloud, no LLMs, no privacy compromise. Real-time speech is cleaned, corrected, and pasted to your cursor in under 1500ms.

Quick Start

Option 1: Download ZIP (Recommended)

1. Download the latest ZIP
2. Extract the folder
3. Open the Speech2Text2 folder
4. Double-click Start.bat — that's it!

The launcher will automatically:

• Create a Python virtual environment
• Install all dependencies
• Start the backend server
• Launch the Electron desktop app

Option 2: Clone with Git

git clone https://github.com/ImAust1n/YapYap.git
cd YapYap/Speech2Text2
Start.bat

Batch Scripts

Script	Purpose
Start.bat	Installs dependencies and launches the app
Stop.bat	Stops the backend server and closes all windows
Update.bat	Pulls latest code and updates dependencies
Uninstall.bat	Removes .venv, node_modules, models, and logs

How to Use

1. Run Start.bat — the app will appear in your system tray
2. Press Ctrl+Shift+D anywhere to activate dictation
3. Speak — text appears live at your cursor
4. Press the shortcut again or wait for the session to end

Architecture

• Backend (backend/app.py) — FastAPI server running the STT + NLP pipeline
• Frontend (electron-app/) — Electron overlay app with global hotkey
• Pipeline: Audio → Whisper STT → Filler Removal → Repetition Detection → Grammar Correction → Tone Control → Formatting → Paste

Requirements

• Windows 10/11
• Python 3.10+
• Node.js 18+
• Git

All Python and Node dependencies are installed automatically by Start.bat.

Development

# Backend
cd backend
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
python app.py

# Frontend (separate terminal)
cd electron-app
npm install
npm start