MiniMax Launches H3 Video Generation Model with Audio and Multimodal Inputs

linoy_tsaban · x · 2026-08-03

MiniMax has released H3, a new 33B parameter video generation model. It features state-of-the-art video generation with synchronized audio and supports text, image, video, and audio references. The model is also optimized to run on consumer GPUs via 🧨 diffusers and ComfyUI.

Related event: MiniMax Releases and Open-Sources Omni-Modal H3 Model(28 posts)→

Original post →

More from Multimodal

Multimodal channel →