A lightweight 0.5B parameter text-to-speech model from FunAudioLLM, optimized for efficient and natural voice synthesis.