Ultra-lightweight text-to-speech with just 15M parameters - CPU optimized and high-quality voice synthesis
