Video anchors improved to boost quality and motion in AI videos
Golden Anchor for Video Generation
Computer Vision and Pattern RecognitionMachine Learning
Summary
Generating videos often starts with a key image, called an anchor, that influences the video's look and movement. The authors found that just making these anchors look nicer in pictures doesn’t always make better videos—in fact, it can sometimes make videos worse. They created a new method called Golden Anchor Learning that adjusts the image generator based on how well the resulting videos perform. This method consistently improves video quality and motion in several tested video generation systems.
What this means in practice
- •For digital content creators: Create higher quality videos with improved motion by optimizing the initial anchor images using video feedback.
- •For game developers: Generate more realistic and dynamic virtual avatars or scenes by refining anchor images that steer video generation models.
- •For advertisers: Produce better AI-driven promotional videos by improving the visual quality and motion through optimized anchor images.$Commercial implications: Enables creation of commercially appealing AI-generated video ads with enhanced realism and motion to attract customers.
Authors
Jinyuan Xu, Shitong Shao, Lichen Bai, Zikai Zhou, Guoxi Huang, Nantheera Anantrasirichai, Zeke Xie
Abstract
Video generation supports digital content creation, with applications in virtual avatars, AI-driven filmmaking, interactive game engines, and world models. State-of-the-art systems achieve controllability primarily through image-to-video (I2V) generation, in which an injected anchor image determines the identity, composition, and style of the resulting video. In modern pipelines, this anchor is synthesized by a text-to-image (T2I) generator. This raises a largely overlooked question: Are images obtained directly from the T2I model truly suitable for current video models? Our analysis shows that anchor aesthetic and quality scores are weakly correlated with downstream video quality. The structure-and-texture score is negatively correlated with I2V video quality in both the visual and motion dimensions. Fine-tuning the T2I model on an image reward accordingly yields anchors with visibly denser structural and character detail: they score higher on image metrics, yet lower the video reward on every I2V model we test and change downstream motion unpredictably. We therefore introduce Golden Anchor Learning (GAL), to our knowledge the first framework for learning golden anchors by optimizing a T2I generator with downstream video feedback. GAL evaluates candidate anchors through the videos they induce under a frozen I2V model and converts video rewards into image-level preferences for DPO fine-tuning of the T2I generator. On all four I2V models, GAL raises the VBench average by $2.97\%$--$8.46\%$ over the untuned generator. The rendered clips also score higher under VideoReward-Overall, and their extracted frames under HPSv3 and UniPercept. GAL yields consistent motion gains across all four I2V models, as measured by a $5.62\%$--$38.24\%$ increase in VBench dynamic degree. Although GAL is trained only on HunyuanVideo-1.5 renders, these gains extend to the other three I2V models.