About the details of the generated images

In the "crystal_ball" and "figurine" demo, the details of the clothes of "figurine" cannot be maintained in the generated image.
Is it possible to solve this problem by finetuning the pretrained model using the custom data?

<img width="1426" height="620" alt="Image" src="https://github.com/user-attachments/assets/390f82fc-1510-4b09-b15a-75511643e320" />