ViTeX-Bench: Benchmarking High-Fidelity Video Scene Text Editing

2026-09-30

AI Summary

This study addresses the persistent challenge in video scene text editing of simultaneously preserving visual fidelity, temporal consistency, and edit locality. To this end, it proposes a systematic solution comprising both evaluation infrastructure and a novel generative model. Methodologically, we construct ViTeX-Bench, a benchmark featuring a real-world paired dataset and an inaugural three-dimensional evaluation protocol. Furthermore, we introduce ViTeX-Edit-14B, an open-source model that integrates OCR calibration, Pareto comparison, and motion-aligned glyph-video conditioned fine-tuning. Experimental results demonstrate that the proposed model achieves a state-of-the-art character accuracy of 0.688 within the field while significantly suppressing text distortion. Ultimately, this work establishes a reproducible research foundation for advancing video text editing.

Abstract

Recent video generation is increasingly realistic and controllable, yet video editing remains less developed, particularly for precise local edits that must preserve the original scene dynamics. Video scene text editing replaces text on scene surfaces, such as storefront signs, whiteboards, and product labels, while preserving the surrounding content, motion, and camera dynamics. Although scene text editing is well studied for images, video scene text editing that achieves high visual quality, temporal consistency, and edit locality remains underexplored. Existing resources offer limited paired real-video data, and general video-editing metrics do not directly measure whether the requested text remains correct over time. We introduce ViTeX-Bench, a benchmark suite comprising ViTeX-Dataset and a three-axis evaluation protocol. The dataset contains 387 real-world 720p videos with text-region masks and editing instructions: 230 provide reviewed, pipeline-generated paired edits for training, and 157 form a frozen evaluation split. The protocol evaluates text correctness, visual and temporal quality, and edit locality through 13 metrics, with one primary metric per axis and a Pareto comparison of their trade-offs. OCR calibration, human evaluation, and annotation-sensitivity analyses support the interpretation of these scores. Across eight baselines from four editing families, accurate text, temporal stability, and scene preservation remain difficult to achieve together. We also release ViTeX-Edit-14B, an open-source reference editor fine-tuned on the paired training split with motion-aligned glyph-video conditioning. It achieves CharAcc 0.688, the highest mean among the evaluated video-native editors, and the lowest comparable text-crop Warp among raw editor outputs. ViTeX-Bench provides a reproducible foundation for studying these trade-offs in video scene text editing.

Problem

• Video Scene Text Editing
• Temporal Consistency
• Edit Locality
• Benchmark Evaluation

Innovation

• Video Scene Text Editing
• Benchmark Suite
• Motion-Aligned Glyph-Video Conditioning
• Three-Axis Evaluation Protocol
• Temporal Consistency