Researchers at the Montreal Institute for Learning Algorithms (MILA) recently OSS’ed SpeechBrain to accelerate the development of neural speech processing technologies by providing a single, efficient, flexible library for multiple speech processing tasks (e.g. speech recognition, spoken language understanding, speaker recognition, and others). Designed to be simple and user-friendly, SpeechBrain includes tools for managing and accelerating training and evaluation loops; optimizing hyperparameters; and building data pipelines. The creators have also released several pretrained models on HuggingFace.