Model description
Model Overview This is a versatile LoRA designed to enable any model to cosplay any character. By leveraging this LoRA, you can seamlessly transfer the outfit and style from an anime character image onto a real-life human subject.
Try It Online Experience the full potential of this model via the optimized workflow on RunningHub: 👉https://www.runninghub.ai/post/2028391084245655554/?inviteCode=sdhs0trb
How to Use Workflow: Use the Qwen Image Edit 2511 workflow.
Prompt: Input: "Make the person in Image 1 cosplay the character in Image 2."
Inputs: * Upload one Anime Character image.
Upload one Real Human Model image.
Important Note: For best results, ensure both images share the same framing. If your model is shown in a full-body shot, the anime character should also be full-body. Otherwise, the AI will have to "hallucinate" or guess the missing parts.
Pro Tips For complex scenes, intricate costumes, or highly stylized/abstract anime references, we recommend using this in conjunction with the Z-Image Redraw (Hires. fix) workflow to enhance details and overall texture.
Description
FAQ
Comments (7)
用gguf和lightlora衣服复制了，但是衣服是糊的，脸是清晰的
In this case, we can only redraw this image using Z image. You can download the workflow from my link. Due to the relatively complex functionality, it exceeds the upper limit of the model's understanding. Sometimes the results are good, and sometimes they are just so-so
I'm using your workflow but there are some lora with non-English descriptions
Actually, whether the cue word is in English or Chinese, as long as the meaning is the same, it can still be understood.
真的是我用过效果最好的Lora之一，非常强大，换装非常自然，谢谢你的分享！
Can you please tell me, what is this Lora?
Z-image绝对真人像写真.safetensors
用2511和lightx2v ,Strength 1 脸和衣服会有瑕疵，降到0.7效果好一点但还是会有一点。大佬用bf16完全不会糊吗？
试着用zimage从画，衣服会偏离原图。从画是不是要用llm形容原图？