A high-performance, local-first text-to-speech engine optimized for consumer hardware.
Local Lab provides a seamless, private interface for generating natural-sounding narration using state-of-the-art neural TTS models. It operates entirely on-device, ensuring zero data leakage and extreme low-latency performance.
The engine provides two distinct operating modes, verified for real-time performance on mid-range consumer hardware.
| Mode | Model Architecture | Real-Time Factor (RTF) | Accuracy | Device |
|---|---|---|---|---|
| Standard | Kokoro-82M (v0.1) | ~0.08 (12.5x faster than real-time) | High | CPU/GPU |
| Pro | StyleTTS 2 (Hardened) | ~0.42 (2.4x faster than real-time) | Premium | NVIDIA |
| Feature | Description |
|---|---|
| 🎤 20+ Voices | American, British, Hindi accents |
| ⚡ Standard Mode | Fast, works on any PC (Kokoro-82M) |
| 🚀 Pro Mode | Premium quality for NVIDIA GPUs (StyleTTS 2) |
| 📥 Export | Download as WAV or MP3 |
| 🎚️ Speed Control | 0.5x to 2.0x |
| 💻 100% Local | No internet needed after setup |
| 🔒 Privacy | Your text never leaves your computer |
- ✅ Windows 10/11
- ✅ Python 3.9+ (Download)
- ⭐ NVIDIA GPU (optional, for Pro Mode)
- Download → Click "Code" → "Download ZIP"
- Extract the ZIP to a folder
- Run
setup.bat(first time only, installs dependencies) - Run
start.batto launch the app - Open http://localhost:8000 in your browser
- ✏️ Type or paste your script
- 🎤 Select a voice (hover to preview!)
- 🎚️ Adjust speed if needed
- 🎬 Click Generate Voice
- 📥 Download your audio!
- Use
...for dramatic pauses - Keep sentences short for better flow
- Use proper punctuation for correct intonation
| Problem | Solution |
|---|---|
| "Python not found" | Install Python 3.9+ from python.org |
| "No module named X" | Run setup.bat again |
| Pro mode locked 🔒 | Install NVIDIA GPU drivers |
Like this tool? Subscribe for more free AI apps!
- Kokoro TTS - Standard Mode
- StyleTTS 2 - Pro Mode
Made with ❤️ by Local Lab
⭐ Star this repo if you found it useful!