I'm a big fan of vintage posters. In the past I've tried to make vintage posters by scraping a ton of vintage posters, together with landscape pictures, and train a CycleGAN in to apply the style of vintage posters to the pictures. The results were underwhelming and training the networks capable of generating images of decent quality required heavy compute. So I gave up on my dreams and moved on with live. Until I came across Domain Enhanced Arbitrary Image Style Transfer via Contrastive Learning (CAST)! This is a Streamlit deployment of this great work.