Inference-time Trajectory Optimization for Manga Image Editing

TL;DR

This paper introduces an inference-time adaptation technique that improves pretrained image editing models for manga images by minimally adjusting the generation trajectory, achieving better fidelity without retraining.

Contribution

The method enables manga-specific image editing by small, inference-time corrections, avoiding costly retraining or fine-tuning of large models.

Findings

Outperforms existing baselines in manga image editing tasks.

Maintains high fidelity with negligible computational overhead.

Effectively adapts pretrained models to manga images without retraining.

Abstract

We present an inference-time adaptation method that tailors a pretrained image editing model to each input manga image using only the input image itself. Despite recent progress in pretrained image editing, such models often underperform on manga because they are trained predominantly on natural-image data. Re-training or fine-tuning large-scale models on manga is, however, generally impractical due to both computational cost and copyright constraints. To address this issue, our method slightly corrects the generation trajectory at inference time so that the input image can be reconstructed more faithfully under an empty prompt. Experimental results show that our method consistently outperforms existing baselines while incurring only negligible computational overhead.