By skyforbes 
Hey everyone! đź‘‹
I've been exploring the new 3.0 model extensively over the past few days (via Vertex AI) and decided to stress-test it focusing on facial consistency + physical interaction.
It started simple: inserting myself into TV shows, but I wanted a bigger challenge regarding lighting and composition. So, I jumped into the MCU.
The main goal wasn't just to 'be there', but to interact. Most models fail when bodies touch (merged hands, bad physics). My prompts focused on forcing the model to understand weight (Wolverine's arm on my shoulder), traction (Deadpool pulling me), and complex lighting.
The results blew me away, especially the skin texture interacting with CGI characters.
Here are the best 19 shots from this experiment. Feedback is welcome!