MangaNinja is a reference-guided line art coloring method that ensures precise transcription of character details through its unique design. This includes a block-shuffling module that facilitates corresponding learning between reference colored images and target line art, as well as a point-driven control scheme that achieves fine-grained color matching. The model has shown outstanding performance in self-collected benchmarks, surpassing existing solutions' coloring accuracy. Furthermore, its interactive point control demonstrates great potential when dealing with complex scenarios such as extreme poses, shadows, and cross-character coloring, challenges that current algorithms struggle with. MangaNinja was co-developed by researchers from the University of Hong Kong, Hong Kong University of Science and Technology, Tongyi Lab, and Ant Group, with the related paper published on arXiv and the code open-sourced.