Users can run inference on their models, train Nvidia Edify foundation models on proprietary data, or start from pre-trained models to generate image, video, and 3d models from text and image prompts.