Identifying the speaker of each dialogue instance is a key step in manga understanding, supporting applications such as translation, summarization, and voice generation. In zero-shot settings, the task is challenging because each new manga title introduces an unseen cast and provides no speaker annotations, predefined character identities, or per-title model adaptation. Existing methods, including iterative multimodal fusion, still organize speaker reasoning around manga-specific visual evidence such as characters’ appearance and local text-character relationships. However, such evidence is highly variable in manga, where speakers may be off-panel, character appearance changes across poses and styles, and local visual relations can be misleading. We propose Speaker-Seeded Identity Consolidation (SSIC), which reframes zero-shot manga speaker attribution as book-level identity consolidation driven by recurring dialogue consistency rather than local visual-region propagation. SSIC induces local anonymous speaker hypotheses from ordered dialogue context, aligns them into cross-window speaker seeds, constructs a global speaker identity registry, and performs registry-constrained attribution. In the main controlled evaluation on Manga109Dialog, SSIC achieves 74.62% speaker attribution accuracy, improving over iterative multimodal fusion by 7.02 points. The gains are largest on hard cases where local text-character relationships are unreliable, and ablations confirm that cross-window speaker seeds are central to registry construction.

End-to-end manga generation is a structured visual storytelling task that requires story decomposition, recurring character and scene grounding, page layout design, panel rendering, page composition, and lettering. However, existing generative models often perform direct page synthesis, entangling these factors in a single visual output and limiting precise control over layout geometry, visual references, and cross-panel consistency. To address these limitations, we propose MangaFlow, an agentic framework for controllable long-form manga generation that decomposes manga creation into planning, grounding, layout construction, reference-conditioned rendering, composition, and text placement. By treating layout and visual references as explicit intermediate variables, MangaFlow enables both simple text-to-manga generation and more precise user-controlled manga creation. This design exposes layout, visual assets, and lettering as editable intermediate controls for refining panel geometry, references, and text placement. To support long-form consistency, MangaFlow introduces a story section memory that links section descriptions with corresponding character, scene, and object references for reuse across panels. We further present a meta-benchmark for evaluating layout controllability, visual consistency, and generation quality. Experiments show that MangaFlow improves layout adherence and cross-panel consistency over direct generation baselines while supporting flexible human control.