Stable Diffusion - CompVis (original research code)
Original latent-diffusion text-to-image implementation
The original Stable Diffusion implementation and research code from CompVis (LMU Munich & Runway). For those interested in generative image models: latent diffusion architecture, conditioning mechanisms, samplers, and how text-to-image generation actually works under the hood. Research-grade - read alongside the paper.
Learn from ReposFreeadvanced~25h
Submitted by @yahianaimRoadmap (3 steps)
1
Concepts
Paper + code reading#Diffusion models intuition#Latent vs pixel space#Conditioning (CLIP)
2
Hands-on
Experiments#Running inference locally#Samplers & schedulers#img2img & inpainting
3
Beyond
Extension#Fine-tuning concepts#LoRA & DreamBooth landscape#Ethics & licensing