Saved in:
Bibliographic Details
Main Authors: Li, Shangxun, Uh, Youngjung
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.16443
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912794757038080
author Li, Shangxun
Uh, Youngjung
author_facet Li, Shangxun
Uh, Youngjung
contents Text-to-image diffusion models excel at generating high-quality images from natural language descriptions but often fail to preserve subject consistency across multiple outputs, limiting their use in visual storytelling. Existing approaches rely on model fine-tuning or image conditioning, which are computationally expensive and require per-subject optimization. 1Prompt1Story, a training-free approach, concatenates all scene descriptions into a single prompt and rescales token embeddings, but it suffers from semantic leakage, where embeddings across frames become entangled, causing text misalignment. In this paper, we propose a simple yet effective training-free approach that addresses semantic entanglement from a geometric perspective by refining text embeddings to suppress unwanted semantics. Extensive experiments prove that our approach significantly improves both subject consistency and text alignment over existing baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2512_16443
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Geometric Disentanglement of Text Embeddings for Subject-Consistent Text-to-Image Generation using A Single Prompt
Li, Shangxun
Uh, Youngjung
Computer Vision and Pattern Recognition
Text-to-image diffusion models excel at generating high-quality images from natural language descriptions but often fail to preserve subject consistency across multiple outputs, limiting their use in visual storytelling. Existing approaches rely on model fine-tuning or image conditioning, which are computationally expensive and require per-subject optimization. 1Prompt1Story, a training-free approach, concatenates all scene descriptions into a single prompt and rescales token embeddings, but it suffers from semantic leakage, where embeddings across frames become entangled, causing text misalignment. In this paper, we propose a simple yet effective training-free approach that addresses semantic entanglement from a geometric perspective by refining text embeddings to suppress unwanted semantics. Extensive experiments prove that our approach significantly improves both subject consistency and text alignment over existing baselines.
title Geometric Disentanglement of Text Embeddings for Subject-Consistent Text-to-Image Generation using A Single Prompt
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2512.16443