First Hands-on with MiniMax H3: Supports 12 References for Video Generation

TomLikesRobots · x · 2026-07-30

Reviewer @TomLikesRobots shared their first generation using the new MiniMax H3 video model. The omni model accepts up to 12 references (audio or video, up to 15 seconds each) and outputs up to 15 seconds of 2K resolution video.

Related event: MiniMax H3 Tested for Native Multimodal Generation(2 posts)→

Original post →

More from Multimodal

Multimodal channel →