xzf-thu/BreezeTTS2_Mac_Streaming

BreezeTTS2_Mac_Streaming provides real-time text-to-speech inference on Apple Silicon using mlx-audio. The project optimizes depth decoders to reduce real-time factors from four to one. It implements intra-frame KV reuse and optional compiled paths for lower latency. Developers can install the Python package to run streaming audio generation locally.