Our third century deals AI agents for image and video generation built in partnership with Google taught by Katie Newman and Wafa Bakelli. You might
have worked with AI agents that produce text output, but human communication includes images, video, and audio. If you are writing a research report
or building a product demo or creating an explain a video, you're probably dealing with visual media.
I use Google's Nanu Bananu image generation and VEO video generation a lot and have been impressed with how quickly these capabilities are improving.
And incorporating these into agentic workflows will let you do even more than the consumer versions of these products. In this course, you build AI
agents that generate images and videos and evaluate the output automatically and iterate until they
meet your quality standards. It turns out a key in building a media agent is in the evaluation or the evals. The core challenge is when the model
generates an image or video from a prompt, there's no single correct answer to compare against [music] and so the quality depends on the context, the
brand requirements, and on your use case. That's right, Andrew. You will learn three evaluation
techniques that we work together. First, you will use fast image text models to calculate the similarity between the image outputs and your text
prompt. Then you will also use an LLM based judge to evaluate against custom criteria like brand consistency or visual quality. This will give you
actionable feedback you can feed back into the next generation. And finally, you will also use structured
rubrics to break down complex prompts into verifiable questions. For example, is the subject in the frame? Does the camera motion match what you asked
for? Or is the style consistent across scenes? If you're ready to build agents [music] beyond text, this course will get you there. I hope you enjoy
the course.