MiniMax-H3 is a 33B parameter state of the art video generation model that produces video and a
fully synchronized soundtrack (ambience, foley, speech).
Turbo mode (optional) applies the larryvrh distillation LoRA
at generation time, reducing inference from 25 steps to 7 for ~4× faster generation with minimal quality loss.
FlashGen mode (optional) applies the Beidouqixing 4-step flashgen LoRA
at generation time, reducing inference from 25 steps to 4 for even faster generation. It ships in the original MiniMax-H3
format and is converted to diffusers layout in-memory at startup.
Apply larryvrh distillation LoRA, reduce steps from 25 to 7
Apply Beidouqixing 4-step distillation LoRA, reduce steps from 25 to 4