Speech To Text (Nemotron)
Introduction · Tutorial · Series: Operators · 6 min 56 sec · Recorded on Composer R3 2026
Overview
The Nemotron speech-to-text operator writes live captions while the speaker is still talking. It runs NVIDIA's speech model on your own GPU, so no audio ever leaves the machine. In this video we add it to an interview, load the model, and watch the captions roll up as the interview plays. Then we tune how quickly they follow the speech and where sentences end, and change how they move on screen. Let's get started.
Keywords
Speech To Text (Nemotron), live captions, speech recognition, NVIDIA Nemotron, Chunk Size, End-of-Sentence Silence, Roll-Up Time, Text Replacement, Transcription, script callback