// ✦ Diffusion, Flow Matching & Generative Models

Diffusion Transformer (DiT)

Replaces the U-Net denoiser with a Transformer.

扩散 Transformer DiT 2022 Advanced core

IN PLAIN WORDS

Patchify latents into tokens and denoise with a Transformer, injecting timestep and conditions with adaptive LayerNorm. It scales well; Sora, SD3 and FLUX use DiT-style designs.

Used in: Sora, Stable Diffusion 3, FLUX, video models

TIMELINE · prerequisites → this → what builds on it

2020 Vision Transf…2021 Latent Diffus…2022 Diffusion Transfo…

01Original paper

Scalable Diffusion Models with Transformers

William Peebles, Saining Xie · 2022

arXiv ↗ ✓ Verified · arXiv API

02Where it sits

Learn the left column first; the right column builds on this one.

Learn first

Diffusion Transformer (DiT)

Then you can learn

Nothing in the atlas builds on it yet

Related:Flow Matching

Left: prerequisites · right: builds on this · dashed: related. Full map →

03Learning resources

Official docs, explainers, courses and reference code. All links checked.

Reference code

github.com ↗EN

Explainers

04Where the course teaches it

U72 Diffusion, Flow Matching & Image Generation VAE/GAN, DDPM/DDIM, score SDEs, latent diffusion, CFG, DiT, flow matching and ControlNet. 14 h15 algorithms6 papers U73 Speech, Video & World Models Whisper, video diffusion (Sora-style DiT), 3-D reconstruction, world models, Dreamer and Genie. 8 h10 algorithms4 papers

On learning paths:Generative Models & Frontier Research

For AI / data

Verification: verified, checked 2026-10-21.
JSON: /api/learn/algorithms.json (id: dit)