benchmark for trs and stt
A high-throughput acoustic pipeline benchmarking local neural TTS (Kokoro) and STT (Whisper) on hardware accelerators, evaluated via the proprietary QVV semantic metric.
