This video demonstrates a real-time text-to-speech application implemented on an Intel® Stratix® 10 NX FPGA. The application implements WaveNet, a deep neural network that produces state-of-the-art audio quality, on the FPGA. It utilizes the Intel® Stratix® 10 NX FPGA’s AI Tensor Blocks, High Bandwidth Memory (HBM) and Block Floating Point 16 (Block FP16) precision to demonstrate 8x-16x increase in throughput compared to the Nvidia V100 GPU.