Stable Diffusion is one of most popular image models available. What makes it different from others such as "MidJourney" is that it’s entirely open-source. StabilityAI, the commercial team behind the model has been working tirelessly to improve its commercial and academic use cases as a foundational model. In this training session, we looked at the specs of Stable Diffusion, and demonstrate techniques which enable custom generation beyond text prompts, such as style-driven generation, subject-driven generation, pose-driven generation; those techniques have collectively empowered many apps and projects.
The presentation deck is available here. This event was hosted by the amazing DSIGN, a branding and communication agency based in Central, Hong Kong.