Lecture 17 – DALL-E 2 and Stable Diffusion
Notes
Recording
Readings
- Foster, Chapter 13
- GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models, Nichol et al., ICML 22
- Hierarchical Text-Conditional Image Generation with CLIP Latents (a.k.a. DALLE-2), Ramesh et al., 2022
- High-Resolution Image Synthesis with Latent Diffusion Models (basis for Stable Diffusion), Rombach et al., CVPR 2022
Related Links
- How DALL-E 2 Works (Aditya Ramesh)
- How DALL-E 2 Actually Works (Ryan O’Connor)
- How does DALL-E 2 Work? (Aditya Singh)
- How is it so good? (DALL-E Explained Pt. 2) (Charlie Snell)
- DALL-E 2 Pre-Training Mitigations (OpenAI)
- DALL-E 2 Creates Incredible Images — and Biased Ones You Don’t See (Wired)
- Stable Diffusion class-action lawsuit
- Have I Been Trained?
- Glaze: Protecting Artists from Style Mimicry by Text-to-Image Models, Shan et al., 2023