ZhipuQingyingIt is the latest product launched by Zhipu AIAI VideoThe generation tool, Zhipu Qingying supports two modes: text-generated video and image-generated video. It can quickly generate 6 seconds of video content with a resolution of up to 1440x960 in just 30 seconds.
Zhipu Qingying is built on a multimodal large model technology base. Its core video generation model is CogVideoX, which uses a new DiT architecture that can more efficiently compress video information and integrate text, time and space dimensions. In addition, Zhipu Qingying also has a high degree of controllability, and users can control the generated video content through language.
Clear Shadow Function
Text-generated video: Users input text descriptions or scripts, and AI generates video content based on these text information.
Image-to-video: Users upload one or more static images, and AI generates dynamic videos after analyzing the image content.
Add background music: Support adding background music to the generated video, including quiet, lighthearted, sad, epic, funny and other types.
Fast generation: Compared with other similar products on the market, Qingying can significantly shorten the time to generate videos. Users only need to wait about 30 seconds to get a 6-second video. The inference time is compressed from minutes to seconds, greatly improving efficiency.
High controllability: Qingying has improved its ability to follow user instructions through optimization technology and self-developed end-to-end video understanding models, making the generated videos closer to the user's creative intentions. Users can get satisfactory results by describing the scene, lens language and details in detail, reducing the number of "trial and error".
Diverse styles: Users can choose different styles to create, including but not limited to cartoon 3D, black and white, oil painting and film style, to meet the needs of personalized and creative expression.
Available to all: Qingying Intelligent Body is open to all users, whether ordinary users or developers, and can be easily used. Enterprises and developers can also integrate text-to-video and image-to-video capabilities into their own applications through API calls.
Built-in music: Qingying also provides a music function, and user-generated videos can be automatically matched with music, which simplifies the video creation process and allows even users without professional skills to quickly produce colorful video content.
Technical basis: The technical support behind Qingying isZhipu AIThe self-developed video generation model CogVideoX adopts a DiT structure similar to Sora, which can integrate text, time and space information. The optimized model has an inference speed that is 6 times faster than the previous generation.
Zhipu Qingying technology features:
Fast generation: It only takes 30 seconds to generate a 6-second video.
Efficient instruction-following ability: Qingying can accurately understand and execute even complex prompts.
Content coherence: The generated video can better restore the movement process in the physical world.
Flexibility in staging: For example, the camera is able to smoothly follow the three dogs in the frame.
Zhipu Qingying industry applications:
Advertising Production
Film Editing
Short video production
Creative content creation
Official website address:https://chatglm.cn/video