VividFace: High-Quality and Efficient One-Step Diffusion For Video Face Enhancement

2025-09-27

AI Summary

Video facial enhancement (VFE) faces three core challenges: texture modeling distortion, temporal inconsistency, and poor generalization with inefficient inference. To address these, we propose VividFace—a one-stage diffusion-based framework. It innovatively integrates single-step flow matching for efficient denoising, employs joint latent- and pixel-space optimization with randomized switching training to enhance reconstruction fidelity and temporal stability, and leverages multimodal large language models (MLLMs) for high-quality facial video data curation. Built upon the pretrained WANX architecture, VividFace adopts a progressive two-stage training strategy. Extensive experiments demonstrate that VividFace achieves state-of-the-art performance in perceptual quality, identity preservation, and temporal consistency, while significantly accelerating inference speed. To foster community advancement, we publicly release both the model and curated dataset.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Video Face Enhancement (VFE) seeks to reconstruct high-quality facial regions from degraded video sequences, a capability that underpins numerous applications including video conferencing, film restoration, and surveillance. Despite substantial progress in the field, current methods that primarily rely on video super-resolution and generative frameworks continue to face three fundamental challenges: (1) faithfully modeling intricate facial textures while preserving temporal consistency; (2) restricted model generalization due to the lack of high-quality face video training data; and (3) low efficiency caused by repeated denoising steps during inference. To address these challenges, we propose VividFace, a novel and efficient one-step diffusion framework for video face enhancement. Built upon the pretrained WANX video generation model, our method leverages powerful spatiotemporal priors through a single-step flow matching paradigm, enabling direct mapping from degraded inputs to high-quality outputs with significantly reduced inference time. To further boost efficiency, we propose a Joint Latent-Pixel Face-Focused Training strategy that employs stochastic switching between facial region optimization and global reconstruction, providing explicit supervision in both latent and pixel spaces through a progressive two-stage training process. Additionally, we introduce an MLLM-driven data curation pipeline for automated selection of high-quality video face datasets, enhancing model generalization. Extensive experiments demonstrate that VividFace achieves state-of-the-art results in perceptual quality, identity preservation, and temporal stability, while offering practical resources for the research community.

Problem

• Enhancing degraded video facial regions with high quality
• Overcoming temporal inconsistency and limited generalization in face enhancement
• Reducing inefficient multi-step denoising processes in video enhancement

Innovation

• One-step diffusion framework for video face enhancement
• Joint latent-pixel training with stochastic switching strategy
• MLLM-driven data curation pipeline for dataset selection