Video Generation with AI

Chapter 8. From Models to Systems

A Note for Early Release Readers

With Early Release ebooks, you get books in their earliest form—the author’s raw and unedited content as they write—so you can take advantage of these technologies long before the official release of these titles.

This will be the 8th chapter of the final book. Please note that the GitHub repo will be made active later on.

If you’d like to be actively involved in reviewing and commenting on this draft, please reach out to the editor at [e-mail].

Learning Objective: In this chapter, you will learn what changes when a generative AI model becomes a product. By the end of this chapter, you will recognize the eight architectural patterns that distinguish production AI applications from research scripts, see how each is solved differently across LTX Desktop, ComfyUI, and InvokeAI, and have implemented the same custom extension on all three.

Throughout this book we’ve moved through video AI from the inside out: foundations and a first generated video (Chapter 1); the data behind generation (Chapter 2); the model itself across pipelines, training, and fine-tuning (Chapters 3 through 5); understanding video with vision-language models (Chapter 6); and synchronizing audio with cross-modal attention (Chapter 7). By the end of Chapter 7, you could load LTX-2’s 22-billion-parameter checkpoint from Hugging Face, drive it from Python, and reason about why each architectural decision was made.

But a model is not ...