Recently, ByteDance's new AI video model Waver 1.0 was officially released. This is a versatile video generation model with stronger capabilities. Waver 1.0 supports text-to-video and image-to-video conversion, offering users a brand-new creative experience. The model significantly outperforms existing open-source and closed-source models in terms of video generation quality and effects.

image.png

In terms of model performance, Waver 1.0 shows outstanding results on Waver-Bench 1.0 and Hermes Motion Testset. According to manual evaluation, Waver 1.0 demonstrates its superiority in motion quality, visual quality, and prompt following.

image.png

Waver 1.0 also has the ability to generate multi-shot narrative videos. When switching between shots and transitioning through time and space, the model ensures high consistency in the core theme, visual style, and overall atmosphere, ensuring the continuity of the video. In addition, Waver 1.0 supports generating videos up to 10 seconds long, allowing for more complete expression of emotions and actions.

In terms of artistic styles, Waver 1.0 supports the generation of videos in various artistic styles, including ultra-realistic, animation, clay, plush, and more, offering users a rich visual feast. In terms of generating complex movements, Waver 1.0 has shown good performance in sports scenarios, although further improvements are still needed in some complex areas.

Waver 1.0 has also expanded its motion capabilities, applicable to the generation of animal movements, providing users with new creative possibilities. No matter where you come from, you can use Waver 1.0 to realize your creativity and jointly create a better future for artificial intelligence-generated content (AIGC).

Project: https://www.waver.video/

Key Points:

🌟 Waver 1.0 is a powerful all-in-one video generation model that supports text and image to video conversion.

🎨 Supports multiple artistic styles and generates videos up to 10 seconds long, providing rich visual effects.

🏆 Performs better than existing models in motion quality and visual quality, suitable for multi-shot narratives.