No, it's img-Video via a standard local minimax H3 (int8 pruned) and the nvfp4 text encoder. It's just a very detailed promt (I loaded the official promt guides into a local Queen 3.6 heretic and entered several examples as templates, so that simple descriptions could then be converted into a structured promt sheet. And then upscaling and interpolation were applied on top. No lora. Minimax is already good with structured promt at 0.6-0.7 megapixels and 20 steps).