Diffusion Models

This is a collection of resources mostly focused on text-to-image, image-to-image generative models.

MIT IAP 2025 short diffusion courses

Two amazing courses on Diffusion from MIT. The courses feature lecture videos, code assignments, lecture notes and tons of great materials.

• Practical Diffusion
• Introduction to Flow Matching and Diffusion Models

Foundational diffusion papers

Denoising Diffusion Probabilistic Models

DDPM paper started the modern diffusion model revolution. Introduces the mathematical framework and training methodology that underlies most current diffusion models. Essential reading for understanding the fundamentals.

Score-Based Generative Modeling through Stochastic Differential Equations

Improved Denoising Diffusion Probabilistic Models

Key improvements to the original DDPM including learned variance, cosine noise schedule, and importance sampling. These techniques significantly improved sample quality and are now standard in most implementations.

Denoising Diffusion Implicit Models

DDIM introduces deterministic sampling for diffusion models, enabling faster generation and meaningful latent space interpolation. This work opened up many practical applications by reducing sampling time from thousands to tens of steps.

High-Resolution Image Synthesis with Latent Diffusion Models

Introduces Latent Diffusion Models (LDM) which allows more efficient training and inference by avoiding expensive computations in the pixel space. This paper started Stable Diffusion family of models.

Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow

Rectified Flow is a new method that shows both great quality and extremely fast generation capabilities (in just a single/few steps).

Notable papers

• Classifier-Free Diffusion Guidance
• Adding Conditional Control to Text-to-Image Diffusion Models
• Consistency Models
• DiT: Scalable Diffusion Models with Transformers
• Elucidating the Design Space of Diffusion-Based Generative Models
• Scaling Rectified Flow Transformers for High-Resolution Image Synthesis (Stable Diffusion 3 paper)
• Normalizing Flows: An Introduction and Review of Current Methods
• FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space
• Zero-Shot Text-to-Image Generation (DALL-E paper)
• Photorealistic Text-to-Image Diffusion Models with Deep Language Understanding (Imagen paper)
• DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation
• Neural Ordinary Differential Equations
• Variational Diffusion Models
• An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversion
• Normalizing Flows are Capable Generative Models

Tutorials & Explanations

• Sander Dieleman blog posts about different aspects of the Diffusion and generative AI as a whole
• Lil’Log - What are Diffusion Models? is a great overview of the diffusion techniques landscape
• Let Us Flow Together - rectified Flow deep dive, lecture notes and code.
• Diffusion Meets Flow Matching: Two Sides of the Same Coin