This book addresses the growing need for architects and designers to critically engage with artificial intelligence in visual and spatial production. It offers a systematic approach to understanding, directing, and constraining generative models. Combining theoretical insights with technical instructions, the book explains how contemporary AI tools such as latent diffusion models can be employed effectively and responsibly. Rather than relying on AI as an automatic image generator, the book proposes a methodology grounded in control, feedback, and authorship. It introduces practical workflows using platforms - like ComfyUI, MidJourney, Grasshopper and others - to explore constrains, prompt structure, and how procedural logic can be harnessed to produce intentional, coherent results. Intended for architects, educators, researchers, and students, this volume positions AI not as a substitute for design expertise, but as a configurable system that must be shaped by the disciplinary and ethical priorities of architecture.