V-Warper: Appearance-Consistent Video Diffusion Personalization via Value Warping

TL;DR

V-Warper is a training-free framework that enhances appearance consistency in video diffusion personalization by combining coarse image-based adaptation with inference-time appearance refinement, avoiding heavy video finetuning.

Contribution

It introduces a novel, training-free, coarse-to-fine personalization method for video diffusion models that maintains appearance fidelity without large-scale video datasets.

Findings

Significantly improves appearance fidelity in personalized videos.

Maintains prompt alignment and motion dynamics effectively.

Operates efficiently without additional video training.

Abstract

Video personalization aims to generate videos that faithfully reflect a user-provided subject while following a text prompt. However, existing approaches often rely on heavy video-based finetuning or large-scale video datasets, which impose substantial computational cost and are difficult to scale. Furthermore, they still struggle to maintain fine-grained appearance consistency across frames. To address these limitations, we introduce V-Warper, a training-free coarse-to-fine personalization framework for transformer-based video diffusion models. The framework enhances fine-grained identity fidelity without requiring any additional video training. (1) A lightweight coarse appearance adaptation stage leverages only a small set of reference images, which are already required for the task. This step encodes global subject identity through image-only LoRA and subject-embedding adaptation. (2) A inference-time fine appearance injection stage refines visual fidelity by computing semantic correspondences from RoPE-free mid-layer query--key features. These correspondences guide the warping of appearance-rich value representations into semantically aligned regions of the generation process, with masking ensuring spatial reliability. V-Warper significantly improves appearance fidelity while preserving prompt alignment and motion dynamics, and it achieves these gains efficiently without large-scale video finetuning.