Enregistré dans:
Détails bibliographiques
Auteurs principaux: Hu, Jinghao, Zhang, Yuhe, Geng, GuoHua, Li, Kang, Zhang, Han
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2602.21273
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917317398495232
author Hu, Jinghao
Zhang, Yuhe
Geng, GuoHua
Li, Kang
Zhang, Han
author_facet Hu, Jinghao
Zhang, Yuhe
Geng, GuoHua
Li, Kang
Zhang, Han
contents Generating multi-frame, action-rich visual narratives without fine-tuning faces a threefold tension: action text faithfulness, subject identity fidelity, and cross-frame background continuity. We propose StoryTailor, a zero-shot pipeline that runs on a single RTX 4090 (24 GB) and produces temporally coherent, identity-preserving image sequences from a long narrative prompt, per-subject references, and grounding boxes. Three synergistic modules drive the system: Gaussian-Centered Attention (GCA) to dynamically focus on each subject core and ease grounding-box overlaps; Action-Boost Singular Value Reweighting (AB-SVR) to amplify action-related directions in the text embedding space; and Selective Forgetting Cache (SFC) that retains transferable background cues, forgets nonessential history, and selectively surfaces retained cues to build cross-scene semantic ties. Compared with baseline methods, experiments show that CLIP-T improves by up to 10-15%, with DreamSim lower than strong baselines, while CLIP-I stays in a visually acceptable, competitive range. With matched resolution and steps on a 24 GB GPU, inference is faster than FluxKontext. Qualitatively, StoryTailor delivers expressive interactions and evolving yet stable scenes.
format Preprint
id arxiv_https___arxiv_org_abs_2602_21273
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle StoryTailor:A Zero-Shot Pipeline for Action-Rich Multi-Subject Visual Narratives
Hu, Jinghao
Zhang, Yuhe
Geng, GuoHua
Li, Kang
Zhang, Han
Computer Vision and Pattern Recognition
Generating multi-frame, action-rich visual narratives without fine-tuning faces a threefold tension: action text faithfulness, subject identity fidelity, and cross-frame background continuity. We propose StoryTailor, a zero-shot pipeline that runs on a single RTX 4090 (24 GB) and produces temporally coherent, identity-preserving image sequences from a long narrative prompt, per-subject references, and grounding boxes. Three synergistic modules drive the system: Gaussian-Centered Attention (GCA) to dynamically focus on each subject core and ease grounding-box overlaps; Action-Boost Singular Value Reweighting (AB-SVR) to amplify action-related directions in the text embedding space; and Selective Forgetting Cache (SFC) that retains transferable background cues, forgets nonessential history, and selectively surfaces retained cues to build cross-scene semantic ties. Compared with baseline methods, experiments show that CLIP-T improves by up to 10-15%, with DreamSim lower than strong baselines, while CLIP-I stays in a visually acceptable, competitive range. With matched resolution and steps on a 24 GB GPU, inference is faster than FluxKontext. Qualitatively, StoryTailor delivers expressive interactions and evolving yet stable scenes.
title StoryTailor:A Zero-Shot Pipeline for Action-Rich Multi-Subject Visual Narratives
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.21273