MiniMax H3 Max Turbo is a high-throughput image-to-video model from MiniMax, post-trained for stronger prompt adherence and improved visual aesthetics. It is co-optimized with fal’s inference stack to deliver faster generation while maintaining output quality. The model can animate a supplied image as the first frame, use an end frame to guide the destination state, or use both for more controlled transitions. It supports 480p, 768p, and 1080p generation, with configurable video duration and prompt expansion modes.