Yuanbo generated movie posters, but the detail capture isn't quite there yet.

P1 P3 are the results. Except for P1, which is a front-facing shot and retained most of the details, the other posters barely look like me. So there's a limitation with the image generation model—it's hard for 2D to match the multi-dimensional fit. Actually, it could learn from video AI by first creating an individual database, using real-time facial recognition for face customization, or allowing multiple photo uploads to restore higher authenticity of the character.