Towards Visual Text Design Transfer Across Languages

TL;DR

This paper introduces MuST-Bench, a benchmark for evaluating visual text translation across languages, and proposes SIGIL, a new framework that improves multilingual visual text generation without relying on style descriptions.

Contribution

The paper presents MuST-Bench for assessing multimodal style translation and introduces SIGIL, a novel multimodal style translation framework that enhances multilingual visual text generation.

Findings

SIGIL outperforms baselines in style consistency and legibility

Existing models struggle with cross-language visual text translation

MuST-Bench provides a new standard for evaluation in this domain

Abstract

Visual text design plays a critical role in conveying themes, emotions, and atmospheres in multimodal formats such as film posters and album covers. Translating these visual and textual elements across languages extends the concept of translation beyond mere text, requiring the adaptation of aesthetic and stylistic features. To address this, we introduce a novel task of Multimodal Style Translation (MuST-Bench), a benchmark designed to evaluate the ability of visual text generation models to perform translation across different writing systems while preserving design intent. Our initial experiments on MuST-Bench reveal that existing visual text generation models struggle with the proposed task due to the inadequacy of textual descriptions in conveying visual design. In response, we introduce SIGIL, a framework for multimodal style translation that eliminates the need for style descriptions. SIGIL enhances image generation models through three innovations: glyph latent for multilingual settings, pretrained VAEs for stable style guidance, and an OCR model with reinforcement learning feedback for optimizing readable character generation. SIGIL outperforms existing baselines by achieving superior style consistency and legibility while maintaining visual fidelity, setting itself apart from traditional description-based approaches. We release MuST-Bench publicly for broader use and exploration https://huggingface.co/datasets/yejinc/MuST-Bench.