Using clips from the backrooms, we trained our Hunyuan Video Lora (13 hours, H100) using Uniform Frame Extraction at 504x360 resolutionPrompt Template:"a video of KBK backrooms, [prompt] , found footage in the style of KBK"