Search papers, labs, and topics across Lattice.
This paper introduces a novel method for the controlled removal of copyrighted animation characters from images generated by text-to-image diffusion models, addressing significant copyright concerns. By optimizing anchor embeddings with structural and detailed constraints, the approach enables precise intervention during the generation process, achieving high fidelity in the remaining image. Experimental results demonstrate that this method outperforms existing techniques in both erasure effectiveness and image quality, while also allowing for multi-target removal and compatibility with other model modification methods.
Achieving state-of-the-art character erasure without sacrificing image fidelity, this method transforms how we handle copyright in AI-generated content.
The exceptional generation capabilities of text-to-image diffusion models have raised copyright concerns, particularly the unauthorized reproduction of animation characters. Existing concept erasure methods fall short for animation character erasure: model modification methods struggle to identify suitable anchors for diverse, highly distinctive characters; prompt-based steering methods lack fine-grained control for precise intervention. These approaches often yield incomplete erasure and degraded image fidelity, hindering real-world deployment. In this paper, we propose a controllable method operating on the model's continuous textual representation to erase target characters during generation. We optimizes an anchor embedding via structural and detailed constraints to serve as a character surrogate, then replaces target-related embeddings with the anchor via a structure-aware adaptive strategy. Experiments show that our method achieves state-of-the-art erasure effectiveness and image fidelity preservation, while supporting controllable erasure degree, multi-target removal, and model transferability. Moreover, our optimized anchors are plug-and-play with current model modification baselines to improve their erasure performance.
