KGEdit: Ambiguity-Aware Knowledge Graphs for Training-Free Precise Video Generation and Editing

2026-05-28

AI Summary

This work addresses the challenges of semantic ambiguity, incorrect concept binding, and cross-frame inconsistency arising from complex textual instructions in training-free video generation. To this end, the authors propose a high-precision, text-driven video editing method that requires no model retraining. By constructing an ambiguity-aware knowledge graph to parse textual prompts into structured semantics, and integrating a Structured Semantic Injection Module (SSIM) with a Temporal-Aware Semantic Control (TASC) mechanism within a diffusion Transformer architecture, the approach enables fine-grained and temporally coherent video generation. The method significantly outperforms existing techniques in editing accuracy, temporal stability, interactive efficiency, and controllability, achieving, for the first time, high-consistency text-to-video editing without any fine-tuning.

Abstract

In recent years, training-free video generation has progressed remarkably. However, when handling complex textual instructions, existing methods still suffer from semantic ambiguity, incorrect concept binding, and cross-frame inconsistency. To address these issues, we propose KGEdit, a structured semantic control framework for text-to-video (T2V) diffusion models. Specifically, we first construct an ambiguity-aware knowledge graph (AAKG) to disentangle and disambiguate the input prompt, converting it into four types of structured semantics: identity, relation, attribute, and negative constraints. We then design a structured semantic injection module (SSIM) to inject these semantic signals into key layers of the diffusion Transformer, enabling fine-grained semantic control. In addition, we introduce a temporal-aware semantic control (TASC) module that dynamically schedules semantic objectives according to the stage-wise characteristics of the denoising process, further improving semantic alignment and temporal consistency. Experiments show that KGEdit outperforms existing methods in editing precision and temporal stability, while offering higher efficiency and controllability in text-driven interaction scenarios.

Problem

• semantic ambiguity
• concept binding
• temporal consistency
• text-to-video generation
• knowledge graph

Innovation

• ambiguity-aware knowledge graph
• structured semantic control
• training-free video generation
• temporal consistency
• diffusion Transformer