BachVid: Training-Free Video Generation with Consistent Background and Character

TL;DR

BachVid is a training-free method that ensures consistent background and character in generated videos by leveraging DiT's attention mechanism to reuse intermediate features, eliminating the need for reference images or training.

Contribution

It introduces the first training-free approach for consistent video generation that exploits DiT's features to maintain background and character consistency across multiple videos.

Findings

Achieves robust background and character consistency in generated videos.

Does not require reference images or additional training.

Leverages attention mechanisms to reuse intermediate features.

Abstract

Diffusion Transformers (DiTs) have recently driven significant progress in text-to-video (T2V) generation. However, generating multiple videos with consistent characters and backgrounds remains a significant challenge. Existing methods typically rely on reference images or extensive training, and often only address character consistency, leaving background consistency to image-to-video models. We introduce BachVid, the first training-free method that achieves consistent video generation without needing any reference images. Our approach is based on a systematic analysis of DiT's attention mechanism and intermediate features, revealing its ability to extract foreground masks and identify matching points during the denoising process. Our method leverages this finding by first generating an identity video and caching the intermediate variables, and then inject these cached variables into corresponding positions in newly generated videos, ensuring both foreground and background consistency across multiple videos. Experimental results demonstrate that BachVid achieves robust consistency in generated videos without requiring additional training, offering a novel and efficient solution for consistent video generation without relying on reference images or additional training.