In this issue:

• we introduce VQGAN + CLIP architecture;

• we discuss the original VQGAN+CLIP paper; 

• we explore the VQGAN+CLIP implementations. 

Enjoy the learning!  

💡 ML Concept of the Day: VQGAN + CLIP 

In Edge#219, we introduced OpenAI’s CLIP as a method that was able to represent both text and images in the same feature space. OpenAI published CLIP and DALL-E together but open-sourced only the former. Almost immediately after CLIP was made available, AI researchers started playing with the idea of combining it with generative adversarial networks (GANs). That evolution produced one of the most popular text-to-image synthesis in the current ecosystem: VQGAN+CLIP.  

The idea of combining CLIP and GANs is very intuitive. By ...