AudioGen: Textually-Guided Audio Generation

With AudioGen, we have showcased our ability to train AI models to excel in the task of text-to-audio generation. AudioGen is an auto-regressive transformer language model that is specifically designed to produce high-quality audio based on textual descriptions. By providing a textual description of an acoustic scene, our model can generate realistic environmental sounds that accurately reflect the description, incorporating intricate recording conditions and complex contextual details.