Wan 2.1 FusionX T2V test. Only prompts and a LoRA file trained on myself were used. The scenes are dynamic, but the prompt-based facial expressions aren't really good — they're quite boring and repetitive.