Author name: Sanghyeon Lee, Minwoo Lee, Euijin Shin, Kangyeol Kim, Seunghwan Choi, Jaegul Choo

arXiv:2603.27637v1 Announce Type: new
Abstract: We introduce a parameter-efficient adaptation method for panel-aware in-context image generation with pre-trained diffusion transformers. The key idea is to compose learnable, panel-specific orthogonal o…