Text-to-Speech using Kyutai's 1.6B parameter streaming model. Supports French and English. Powered by ZeroGPU.