Test shows Minimax H3 generating 1-minute videos with consistent details without reference images

aziz4ai · x · 2026-09-02

A user shared a test result demonstrating the capabilities of the Minimax H3 model. They successfully generated a video lasting over a minute without using a single reference image, while maintaining highly stable details and coherent scene continuity. The user noted that with clearer character descriptions, consistency could potentially reach 100%. The demonstration was executed on the NitxStudio platform.

Original post →

More from Multimodal

Multimodal channel →