English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Image-to-video generation, which aims to generate a video starting from agiven reference image, has drawn great attention. Existing methods try toextend pre-trained text-guided image diffusion models to image-guided videogeneration models. Nevertheless, these methods often result in either lowfidelity or flickering over time due to their limitation to shallow imageguidance and poor temporal consistency. To tackle these problems, we propose ahigh-fidelity image-to-video generation method by devising a frame retentionbranch based on a pre-trained video diffusion model, named DreamVideo. Insteadof integrating the reference image into the diffusion process at a semanticlevel, our DreamVideo perceives the reference image via convolution layers andconcatenates the features with the noisy latents as model input. By this means,the details of the reference image can be preserved to the greatest extent. Inaddition, by incorporating double-condition classifier-free guidance, a singleimage can be directed to videos of different actions by providing varyingprompt texts. This has significant implications for controllable videogeneration and holds broad application prospects. We conduct comprehensiveexperiments on the public dataset, and both quantitative and qualitativeresults indicate that our method outperforms the state-of-the-art method.Especially for fidelity, our model has a powerful image retention ability anddelivers the best results in UCF101 compared to other image-to-video models toour best knowledge. Also, precise control can be achieved by giving differenttext prompts. Further details and comprehensive results of our model will bepresented in https://anonymous0769.github.io/DreamVideo/.

DreamVideo: High-Fidelity Image-to-Video Generation with Image Retention and Text Guidance