MiniMax H3 to Add 2K Video Module and Faster Local Generation Soon

MiniMax H3 roadmap includes a 2K video module and local execution optimizations. Current 768p limit expands to high-res output, addressing blur issues in multi-modal generation.

Woofun AI reports that the MiniMax H3 team disclosed upcoming enhancements during a Reddit AMA, including the release of a 2K video module and computational optimizations for faster local execution. Currently, H3 supports only 768p video generation locally; the new module will process reference materials within the model to restore fine details in text, characters, and scenes, potentially enabling full 2K workflows locally. The team also acknowledged inherent model limitations where simultaneous image, video, and audio referencing causes blurriness, while distant scenes and small figures often exhibit distorted shapes, noting that further optimizations are underway.