Minimax R2V Workflow: How to Accurately Control Reference Poses

SilliusApeus · reddit · 2026-08-06

A user encountered challenges with reference image control while using the Minimax R2V model for video generation. When providing a start image and multiple pose references, the model rigidly applies the images, causing abrupt environmental changes.

The user tried tagging references as 'weak' and detailing the transition environment, but the results remained random. The post discusses the current limitations of video generation models in complex prompt adherence and multi-subject control, seeking effective practical prompting examples.

Original post →

More from Multimodal

Multimodal channel →