ViSTA: Visual Storytelling using Multi-modal Adapters for Text-to-Image Diffusion Models

TL;DR

ViSTA introduces a multi-modal history adapter for text-to-image diffusion models to generate coherent, narrative-aligned image sequences in visual storytelling, addressing previous limitations in adaptability and consistency.

Contribution

It proposes a novel multi-modal history adapter and salient history selection strategy for improved visual storytelling with diffusion models.

Findings

Achieves coherent and narrative-aligned image sequences.

Outperforms existing methods on StorySalon and FlintStonesSV datasets.

Provides a new metric, TIFA, for assessing text-image alignment.

Abstract

Text-to-image diffusion models have achieved remarkable success, yet generating coherent image sequences for visual storytelling remains challenging. A key challenge is effectively leveraging all previous text-image pairs, referred to as history text-image pairs, which provide contextual information for maintaining consistency across frames. Existing auto-regressive methods condition on all past image-text pairs but require extensive training, while training-free subject-specific approaches ensure consistency but lack adaptability to narrative prompts. To address these limitations, we propose a multi-modal history adapter for text-to-image diffusion models, \textbf{ViSTA}. It consists of (1) a multi-modal history fusion module to extract relevant history features and (2) a history adapter to condition the generation on the extracted relevant features. We also introduce a salient history selection strategy during inference, where the most salient history text-image pair is selected, improving the quality of the conditioning. Furthermore, we propose to employ a Visual Question Answering-based metric TIFA to assess text-image alignment in visual storytelling, providing a more targeted and interpretable assessment of generated images. Evaluated on the StorySalon and FlintStonesSV dataset, our proposed ViSTA model is not only consistent across different frames, but also well-aligned with the narrative text descriptions.