Phillip Isola's lecture from MIT's 6.7960 Deep Learning course covers conditional generative models, the branch of generative modeling where output is shaped by an input signal rather than sampled freely. He works through conditional GANs, conditional VAEs, and conditional diffusion models, explaining how each architecture incorporates conditioning information into training and sampling. The lecture then moves to applications: paired and unpaired image translation, image-to-image generation, text-to-image generation, text-to-text generation, and image-to-text generation, tying each task back to the underlying conditional modeling framework. Running 82 minutes, this is a full lecture from MIT's OpenCourseWare Fall 2024 deep learning course, aimed at students who already have grounding in generative models like GANs, VAEs, and diffusion, and want to see how conditioning extends them into practical translation and synthesis tasks.