Diffusion and Generative Images
How Machines Learned to Draw
- Pages
- 80
- Chapters
- 61
- Level
- Intermediate
- Format
- E-book
About this book
Typing a sentence and receiving a convincing picture still feels astonishing. This book explains what happens in between, and how to use generative image tools responsibly.
You will learn how images are compressed into latent spaces, how earlier generative models such as GANs and VAEs worked, and how diffusion models learn to remove noise step by step. Chapters cover text conditioning and guidance, sampling settings, editing techniques such as inpainting, fine-tuning on custom styles, evaluation, and video generation. The final part addresses copyright, consent, deepfakes, provenance and labelling.
Intuition comes first; the mathematics is kept light and explained in words.
What you will learn
- Latent spaces, autoencoders and earlier generative models
- How diffusion models learn to remove noise
- Text conditioning, guidance and sampling settings
- Editing, inpainting and controlled generation
- Evaluating generated images
- Copyright, consent, deepfakes and provenance
Table of contents
- IntroductionIntroduction
- Chapter 1What Generative Models Do
- Chapter 2Images as Data
- Chapter 3Probability Over Images
- Chapter 4Compression and Latent Spaces
- Chapter 5Autoencoders
- Chapter 6Variational Autoencoders
- Chapter 7Generative Adversarial Networks
- Chapter 8Why GANs Gave Way
- Chapter 9Autoregressive Image Models
- Chapter 10The Diffusion Idea
- Chapter 11The Forward Process: Adding Noise
- Chapter 12The Reverse Process: Learning to Denoise
- Chapter 13Sampling
- Chapter 14Samplers and Steps
- Chapter 15Latent Diffusion
- Chapter 16Text Encoders
- Chapter 17Conditioning With Cross-Attention
- Chapter 18Classifier-Free Guidance
- Chapter 19Negative Prompts
- Chapter 20Resolution and Aspect Ratio
- Chapter 21Prompting Image Models
- Chapter 22What Prompts Cannot Do Well
- Chapter 23Image-to-Image Generation
- Chapter 24Inpainting and Outpainting
- Chapter 25Structural Control
- Chapter 26Image Prompts and Style References
- Chapter 27Fine-Tuning on Custom Concepts
- Chapter 28Consistency Across Images
- Chapter 29Upscaling and Restoration
- Chapter 30Evaluating Generated Images
- Chapter 31Common Artefacts
- Chapter 32Video Generation
- Chapter 333D Generation
- Chapter 34Audio and Music Generation
- Chapter 35Multimodal Models
- Chapter 36Training Data
- Chapter 37Bias and Representation
- Chapter 38Copyright and Ownership
- Chapter 39Consent and Likeness
- Chapter 40Deepfakes and Misinformation
- Chapter 41Provenance and Labelling
- Chapter 42Safety Filters
- Chapter 43Disclosure
- Chapter 44Product Imagery and Advertising
- Chapter 45Design Workflows
- Chapter 46Architecture, Interiors and Product Design
- Chapter 47Education and Science Communication
- Chapter 48Running Models Locally
- Chapter 49Costs and Performance
- Chapter 50An Organisational Policy for Image Generation
- Chapter 51A Denoising Step in Detail
- Chapter 52Prompt Engineering Patterns for Images
- Chapter 53Building an Internal Image Service
- Chapter 54Accessibility and Inclusion in Generated Imagery
- Chapter 55Measuring Value
- Chapter 56Case Study: A Marketing Team
- Chapter 57Case Study: A Publisher's Illustrations
- Chapter 58Case Study: Synthetic Training Data
- Chapter 59How the Field Is Moving
- Chapter 60Mental Models to Keep
- Chapter 61A Project Checklist
- AppendixExercises
- GlossaryGlossary
Who this book is for
Designers, developers and curious readers who want to understand and use image generation thoughtfully.
Details
| Format | PDF + web edition |
|---|---|
| Length | 80 pages (approx.) |
| Chapters | 61 |
| Level | Intermediate |
| Language | English |
| Topic | Deep Learning |
| Author | Latent Library Editorial |
Keep reading