Kuaishou's large video generation model, "KeLing," has recently introduced two new features: "Image to Video" and "Video Continuation," further enriching users' video creation toolkit.

The "Image to Video" feature allows users to upload any image and quickly generate a 5-second精彩视频. This feature not only simplifies the video production process but also supports users in adding prompt words to control the movement of the image, making the generated videos more personalized and creative. Whether it's a merchant looking to showcase dynamic product effects or an ordinary user hoping to create interesting content, this feature makes it easy to bring their ideas to life.

image.png

Here, I tested it with a previously generated image of a beauty under a cherry blossom tree, simply inputting "the camera gradually zooms in, hair and flowers swaying in the wind," and then clicked to generate.

It can be seen that the character videos generated by KeLing have very realistic movements, and the faces are not distorted, appearing very natural.

The "Video Continuation" feature provides the possibility of extending the creation of existing video segments. Users can continue a generated video with a single click for 4 to 5 seconds of content, and support multiple continuations, up to 3 minutes in total. During this process, users can fine-tune prompt words to control the direction and style of the continuation content, ensuring the coherence and consistency of the video content. This feature is especially suitable for scenarios that require long narratives or displays, such as educational explanations and storytelling.

In addition, to better adapt to the needs of different scenarios, "Text to Video" has added two new video size options: 9:16 and 1:1. This means users can choose the appropriate video ratio according to different platform requirements or personal preferences, making the generated content more in line with the characteristics of the publishing platform and enhancing the viewing experience.

Highlights:

1. Powerful Generation Capabilities:

Supports the generation of high-definition videos up to 3 minutes long at 30fps, with resolutions as high as 1080p.

Supports various aspect ratios, including vertical videos, making it very suitable for Kuaishou's short video ecosystem.

2. Motion and Physics Simulation:

Accurately depicts complex and large-scale movements, such as a tiger running on the road, an astronaut walking on the moon, etc.

Can simulate physical characteristics, such as gravity, fluid flow, and optical reflections, making the generated images more in line with real physical laws.

3. Wide Range of Application Scenarios:

From entertainment to education, KeLing's application scenarios are extensive.

Supports text-to-video conversion, multiple control information inputs, and rich content control capabilities.

"KeLing" large model can accurately depict complex and large-scale spatiotemporal movements, such as a tiger running at high speed on the road, with coherent images and coordinated actions, and even the tremors of the torso during the run are meticulously displayed. It can also simulate the characteristics of the real physical world, such as the law of gravity and the rise of the liquid surface when pouring milk, as well as optical reflection laws.

Official Website: https://top.aibase.com/tool/kelingdamoxing