MeDM: Mediating Image Diffusion Models for Video-to-Video Translation with Temporal Correspondence Guidance

TL;DR

MeDM is a novel method that leverages pre-trained image diffusion models for consistent, efficient video-to-video translation and editing, ensuring temporal coherence without fine-tuning or test-time optimization.

Contribution

It introduces a framework that enforces temporal consistency using optical flows and physical constraints, compatible with existing diffusion models, without requiring additional training.

Findings

Achieves high-quality, temporally consistent video translation.

Outperforms existing methods on various benchmarks.

Enables text-guided video editing without fine-tuning.

Abstract

This study introduces an efficient and effective method, MeDM, that utilizes pre-trained image Diffusion Models for video-to-video translation with consistent temporal flow. The proposed framework can render videos from scene position information, such as a normal G-buffer, or perform text-guided editing on videos captured in real-world scenarios. We employ explicit optical flows to construct a practical coding that enforces physical constraints on generated frames and mediates independent frame-wise scores. By leveraging this coding, maintaining temporal consistency in the generated videos can be framed as an optimization problem with a closed-form solution. To ensure compatibility with Stable Diffusion, we also suggest a workaround for modifying observation-space scores in latent Diffusion Models. Notably, MeDM does not require fine-tuning or test-time optimization of the Diffusion Models. Through extensive qualitative, quantitative, and subjective experiments on various benchmarks, the study demonstrates the effectiveness and superiority of the proposed approach. Our project page can be found at https://medm2023.github.io