Running OpenAI’s Whisper Automatic Speech Recognition on a Live Video Transcoding Server

As video streaming workflows increasingly move from file-based to live, functions that previously could be done offline can pose technical challenges or be exceedingly expensive to implement. One such function is the requirement for subtitle creation and transcription for live video streams. In this demonstration, you will see how ten NETINT T1U VPUs installed in a 1RU server with a 96-core Ampere Altra Max processor running OpenAI’s Whisper ASR model can produce dozens of simultaneous ABR ladder transcodes using AV1, HEVC, or H.264.