Introduction to Rectified Flow and Stable Diffusion 3.5

A practical course on modern diffusion models

Learn the key ideas behind Rectified Flow and Stable Diffusion 3.5, then implement the core pieces from scratch. The course covers the essential background, architecture, sampling, and the main components behind modern diffusion systems.

PyTorch

Hands-on code

From Scratch

Theory → Practice

Presale pricing — newsletter and previous students

0d 0h 0m 0s

What You'll Learn

Understand the core ideas behind Rectified Flow and modern diffusion models
Review the key background concepts behind Stable Diffusion pipelines
Learn the role of Latent Diffusion Models, DPM samplers, CLIP, T5, and CFG
Study the main architectural and sampling choices behind Stable Diffusion 3.5
Implement the Rectified Flow paper from scratch
Build the main components of a Stable Diffusion 3.5 step by step

Who This Course Is For

ML engineers who want a deeper understanding of modern image generation systems
Researchers and students reading diffusion and flow-matching papers
Advanced learners who want more than high-level intuition
People comfortable with Python and basic deep learning concepts

Prerequisites

Python programming experience
Basic familiarity with PyTorch
Some prior exposure to deep learning is recommended