Empero releases PiVo Studio Rust TTS server

PiVo Studio is a standalone Rust binary for running the 1.4-million-parameter text-to-speech model PiVo without requiring Python, NumPy, or ONNX Runtime. The engine decodes ONNX graphs using a custom protobuf decoder and hand-written f32 kernels, operating faster than real time while streaming audio within milliseconds and peaking at around 50 MB of memory. It provides an OpenAI-compatible HTTP speech endpoint, a local web studio interface, and a C-compatible library for embedding models directly into mobile applications. Open article →