RGFVR: Reference-Guided Face Video Restoration with Flow Matching
2026-06-15 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors address the challenge of restoring face videos that are blurry, noisy, or compressed while keeping the person's identity clear and consistent over time. They create a new method that uses a special type of video generator guided by both visual and descriptive identity information that doesn't depend on knowing the person beforehand. Their approach is trained in two steps to better preserve details and smoothness in the restored videos. Tests show their method works better than others at fixing different kinds of video problems without losing who the person is.
face video restorationidentity preservationtemporal consistencyflow-based modelstext-to-video generationreference-guided restorationvideo degradationperceptual identity conditioningsubject-agnostic modeltwo-stage training
Authors
Cem Eteke, Batuhan Tosun, Eckehard Steinbach
Abstract
Face video restoration from degraded observations is challenging, as it requires simultaneously recovering visual fidelity, temporal consistency, and subject identity. Existing approaches are often either reference-free, which can lead to identity loss when person-specific facial details are lost, or subject-specific, which limits generalization to unseen identities. We propose a subject-agnostic, reference-guided framework for identity-preserving face video restoration. Our method introduces bimodal perceptual-descriptive identity conditioning into a pretrained flow-based text-to-video generator and employs a two-stage training strategy to strengthen identity guidance during restoration. Experiments show that our approach improves restoration fidelity, temporal consistency, and identity preservation, achieving superior performance under challenging video degradations, including downsampling, blur, noise, and compression artifacts. The code is available under: https://github.com/batuhanntosun/RG-FVR.