Best way to use H3 ref2video. Not even joking. it just works and is easier to manage than any replacing attempts fighting the Model
retained text
Just draw your composition and let Minimax figure it out instead of using references from the internet Side Note: I noticed when trying to replace people, the Model struggles a lot when it is physically not straightforward to fit something in. Like when your hair has a tied bun behind your head, Minimax struggles to lie you on the floor. You have to tell Minimax to untie the hair. Which gets really difficult when it is very complex with a lot of minor conflicts between the original and replacement subject
document fetched 2026-10-12 00:29 · canonical https://reddit.com/r/StableDiffusion/comments/1x3c4vm/best_way_to_use_h3_ref2video_not_even_joking_it