This book is licensed for the personal use of the buyer. Please do not share or resell the file. Short quotations with attribution are welcome.
The information in this book is provided for education. It is not legal, financial or other professional advice. Product and company names are used only to describe them and belong to their owners.
This is a free sample: the introduction and first chapter.
IntroductionIntroduction
Type a sentence and receive a picture that never existed: a lighthouse in the style of a watercolour, a product photo on a beach at sunset, an architectural sketch of a library made of glass. A few years ago this was science fiction. Today it is a feature in design tools, phones and office software.
This book explains what happens between the sentence and the picture, and how to use image generation thoughtfully. It starts with how images are represented and compressed, surveys earlier generative models, and then explains diffusion models step by step: adding noise, learning to remove it, guiding the process with text, and sampling. Later chapters cover editing, control, fine-tuning on custom styles, evaluation, video, and the questions of copyright, consent, deepfakes and provenance that every user and organisation must take seriously.
Intuition comes first. The mathematics is kept light and always explained in words.
How the book is organised
Chapters 1 to 9: images, latent spaces, autoencoders, GANs and VAEs.
Chapters 10 to 20: how diffusion works: noise, denoising, sampling, latent diffusion, text conditioning and guidance.
Chapters 21 to 31: using and controlling generation: prompting, editing, control, fine-tuning and evaluation.
Chapters 32 to 40: video, 3D, audio and multimodal generation.
Chapters 41 to 50: provenance, safety, disclosure, uses across fields, local models, costs and policy.
Chapters 51 to 55: a denoising step in detail, prompt patterns, an internal image service, accessibility and measuring value.
Chapters 56 to 61: case studies, where the field is going, mental models and a checklist.
Chapter 1What Generative Models Do
A generative model learns the patterns of a collection of data well enough to produce new examples that resemble it. A classifier answers "what is this?"; a generator answers "show me something like this".
Generative image models can:
create images from text descriptions;
vary or extend existing images;
fill in missing regions;
change style while keeping content;
increase resolution.
You have reached the end of the free sample. The full book continues with 60 more chapters, plus the appendices.