LoRA Support

Models that accept a LoRA fine-tune at inference time — load your brand-trained weights, apply a custom style or character, stay consistent across a long output run. Today's catalog spans Qwen Image Edit Plus LoRA, Qwen Image Edit 2511 LoRA, and the Wan 2.2 i2v LoRA variants (fast + ultra-fast). LoRA is always a secondary capability — the model's primary modality tells you what it generates.