| title | Whisper WebGPU + Llama Chat |
|---|---|
| emoji | 🎙️🦙 |
| colorFrom | blue |
| colorTo | green |
| sdk | docker |
| pinned | false |
| app_port | 7860 |
# 🎙️ Whisper Web + Llama 3.2 🦙
ML-powered speech recognition and AI chat, right in your browser
Features • Quick Start • Models • Development • Tech Stack
- 🎤 Real-time Processing - Instant speech recognition with Whisper and AI responses with Llama 3.2 1B Instruct
- 🌐 100% Client-side - No server needed, everything runs in your browser
- ⚡ WebGPU Powered - Hardware-accelerated for optimal performance
- 🔒 Privacy First - All processing happens locally
- 🌍 Multilingual - Supports multiple languages
# Clone the repository
git clone https://github.com/Reidmen/whisper-llama-web.git
# Install dependencies
pnpm install
# Start the development server
pnpm run devVisit http://localhost:5173 in your browser.
| Model | Size | Languages | Performance |
|---|---|---|---|
| Tiny | 120MB | All | Fast |
| Base | 206MB | All | Balanced |
| Small | 586MB | All | Accurate |
| Large V3 | 1.6GB | All | Most Accurate |
| Distil Small | 538MB | English | Optimized |
- Llama 3.2 1B Instruct (Quantized for browser)
- Fast local inference
- ~1GB memory usage
- Multi-turn conversations
- Modern browser with WebGPU support
- GPU with sufficient memory
- Microphone access (for recording)
- Clone and install dependencies
- Configure environment variables (if needed)
- Start development server
- Open browser at localhost:5173
docker build -t whisper-web .
docker run -p 7860:7860 whisper-web- Frontend: React, TailwindCSS
- ML Framework: Transformers.js
- Performance: WebGPU, Web Workers
- Build Tools: Vite, TypeScript
MIT License - see the LICENSE file for details
Built by Reidmen, inspired by the transformers.js