Subject
7 entries
Diffusion Models
Bookmarks
The breakthrough behind modern AI image generators: diffusion models (3Blue1Brown)
3Blue1Brown's visual explanation of diffusion models — part 1 covering the core breakthrough behind modern AI image generators. Classic 3B1B treatment: intuition-first, mathematically grounded, visually clear.
ControlNet: Precise Spatial Control for Diffusion Models
ControlNet adds fine-grained spatial control to Stable Diffusion — use edge maps, depth maps, pose skeletons, or sketches to precisely direct where objects and structures appear in generated images. A major step beyond text-only prompting for image generation.
Human Motion Diffusion Model (MDM)
Tevet et al. (Tel Aviv University, arXiv:2209.14916, 2022) apply diffusion models to human motion generation, producing MDM — a transformer-based denoiser that generates realistic motion sequences from text descriptions or action labels. It matters because it extends the generative power of diffusion to a structured temporal domain, enabling controllable motion editing that prior methods couldn't match.
DIFFUSER: Discrete Diffusion via Edit-Based Reconstruction
Introduces DIFFUSER, an edit-based text generation model that adapts denoising diffusion to discrete text by framing generation as iterative editing rather than left-to-right token production. Competitive with autoregressive models on translation and summarization while enabling unique capabilities like prototype-conditioned generation and iterative revision.
An Image is Worth One Word: Personalizing Text-to-Image Generation Using Textual Inversion
Tel Aviv University and NVIDIA paper introducing Textual Inversion — learning a single new text embedding token that represents a user-provided concept, enabling that concept to be composed into any text prompt. Showed that the embedding space of text-to-image models is richly structured and can be expanded with just 3-5 example images.
The Illustrated Stable Diffusion
Jay Alammar's visual explainer of how Stable Diffusion works under the hood — covering latent diffusion, the CLIP text encoder, and the U-Net denoiser. Alammar's illustrated series is one of the best entry points for building intuition about complex ML architectures.
How Diffusion Models Work: The Math from Scratch
AI Summer's mathematical walkthrough of how diffusion models work from scratch — covering the forward noising process, reverse denoising, DDPM training objective, and score matching. The most math-forward accessible introduction to the field.
