Lightweight text-to-speech models that run fast on CPU, edge devices, and low-VRAM GPUs without sacrificing natural voice quality.