I had some extra time this holiday break, so I decided to organize my movie collection. Once I categorized the movies I own into better folders, I realized I am not a fan of the way Plex auto-generates category posters. I decided to design my own.
I wanted something with a dynamic street art, splatter paint aesthetic. Unfortunately I did not practice illustration (digital or analogue) since my university days – so I decided to delegate the labor to an AI.
Here’s a bunch of posters I managed to put together in one evening:
The original generations were created in MidJourney, using a prompt following this general logic:
/imagine prompt poster for the cinematic cinema projector experience, Unreal Engine, Street Art, Artwork, Splatter Paint, Cinematic, Dramatic --ar 2:3 --v 4
Resulting images tend to have nonsensical AI typography on them, so I would then bring them to Adobe Photoshop and use a combination of paint-overs, cloning and content-aware fill to remove the glyphs. (Can I just gush over how amazed I am with content-aware tools in Photoshop? They just keep getting better and better and do in seconds what used to take hours).
Once prepped, I would add new typography on top and then use a variety of brushes and textures from amazing Texturelabs.org to properly grunge it up and unify the look.
Needless to say, I am quite pleased with the end result. It’s just another perfect use case for text-to-image AI generators – a quick image with striking look. Messy aesthetic that hides the deficiencies of artificial intelligence, primed for hand manipulation by a human.
Here’s a bunch of some other images generated in the process:
Big fan of the fact that the keyword Superhero tends to generate Batman with a Superman crest on his costume 🙂