TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text Rendering

TL;DR

This paper introduces TextPecker, a reinforcement learning strategy that improves the structural fidelity of text in generated images by perceptively identifying anomalies, significantly enhancing visual text rendering quality.

Contribution

We develop a novel RL-based method with a character-level anomaly dataset and stroke-editing engine to improve structural accuracy in text-to-image models.

Findings

Achieves 4% improvement in structural fidelity for Chinese text

Yields 8.7% better semantic alignment

Establishes new state-of-the-art in high-fidelity visual text rendering

Abstract

Visual Text Rendering (VTR) remains a critical challenge in text-to-image generation, where even advanced models frequently produce text with structural anomalies such as distortion, blurriness, and misalignment. However, we find that leading MLLMs and specialist OCR models largely fail to perceive these structural anomalies, creating a critical bottleneck for both VTR evaluation and RL-based optimization. As a result, even state-of-the-art generators (e.g., Seedream4.0, Qwen-Image) still struggle to render structurally faithful text. To address this, we propose TextPecker, a plug-and-play structural anomaly perceptive RL strategy that mitigates noisy reward signals and works with any textto-image generator. To enable this capability, we construct a recognition dataset with character-level structural-anomaly annotations and develop a stroke-editing synthesis engine to expand structural-error coverage. Experiments show that TextPecker consistently improves diverse text-to-image models; even on the well-optimized Qwen-Image, it significantly yields average gains of 4% in structural fidelity and 8.7% in semantic alignment for Chinese text rendering, establishing a new state-of-the-art in high-fidelity VTR. Our work fills a gap in VTR optimization, providing a foundational step towards reliable and structural faithful visual text generation.