What is Qingying 2.0
Qingying 2.0 is an upgraded version launched by Zhipu AIAI video generationtool, based onthe CogVideoX modeldeveloped. It supports generating ultra-high-definition videos of 10 seconds, 4K resolution, and 60 frames, and has the following features: model capabilities are comprehensively improved, enabling better understanding of complex text prompts and generation of more realistic videos. It supports video output in any aspect ratio, adapting to different playback needs, and adds a built-in sound effect feature, throughthe CogSound modelto generate matching sound effects for videos, providing multi-channel generation capability, so that multiple videos can be generated at once from the same instruction or image.
Main features of Qingying 2.0
- 4K ultra-high-definition video generation: Supports generating ultra-high-definition videos of 10 seconds duration, 4K resolution, and 60 frames, for a smoother visual experience.
- Any aspect ratio output: Can generate videos in different aspect ratios, adapting to various playback needs.
- Built-in sound effect feature: Through the CogSound model, it can generate matching sound effects based on video content, such as natural environmental sounds, vehicle sounds, and more.
- Multi-channel generation capability: The same instruction or image can generate multiple videos at once, improving creative efficiency.
- Powerful text understanding capability: Can better understand complex text prompts and generate videos that better meet user needs.
- High-quality image-to-video: Supports generating videos by uploading images, further optimizing the precision of character movements and physical world simulation.
- Efficient compression and coherence: Uses a 3D variational autoencoder (3D VAE) to compress video data to 2%, significantly reducing computing resource requirements while maintaining coherence between video frames.
- 3D rotary position embedding (3D RoPE): Improves the model's ability to capture inter-frame relationships in the temporal dimension, establishes long-range dependencies, and generates smoother videos.
Upgrade highlights of Qingying 2.0
- Significant improvement in model capabilities: The model structure, training method, and data engineering have been fully updated, and the basic model capability for image-to-video has increased by 38%. It supports large-scale movement of the main subject in the frame while maintaining image stability, and the motion coherence and richness of camera movement are greatly improved.
- Industry-leading instruction-following capability: Can understand and implement complex prompts, and generate content that better meets user needs.
- Supports multiple artistic styles: Qingying 2.0 can handle realistic styles, 3D animation, 2D animation, and special artistic styles, generating exquisite images.
- Improved visual aesthetics and stability: Performs better in image stability and motion coherence, and the generated videos are more delicate and natural.
- Official examples
- Prompt: The puppy puts the pot down on the gas stove, picks up the spoon on the table and stirs it in the pot
-
- Prompt: Realistic style, family portrait comes to life
-
- Prompt: Viral gold coin special effects
How to use Qingying 2.0
Use through the Zhipu Qingyan web version or APP: Qingying 2.0 is open for free use, with no limit on the number of generations.
- Web version: VisitZhipu Qingyan official website, select the AI video generation page, enter a text description or upload an image to generate a video.
- APP: Download the "智谱清言APP" from major app stores and experience the 清影 feature within the app.
- Usage Recommendations：
- Free Trial: All users can use 智谱清影2.0 for free, though there may be queues.
- Acceleration Service: If you need faster generation speed, you can purchase the acceleration service.
- Parameter Settings: The following parameter modes produce better generation results.
Application Scenarios of 清影2.0
- Education: 清影2.0 can help teachers and educational content creators quickly generate instructional videos. Teachers can input a description about a scientific principle and generate intuitive animated videos to help students better understand complex concepts.
- Marketing: Enterprises can use 清影2.0 to generate customized advertising videos. By inputting product descriptions and marketing copy, 清影2.0 can generate attractive advertising videos and improve promotional effectiveness.
- Entertainment: Content creators can use 清影2.0 to easily produce short videos and movie trailers. 清影2.0 provides rich customization options and efficient video generation capabilities, allowing creators to quickly generate high-quality video works according to their own ideas and needs.
- News Media: News media organizations can use 清影2.0 to quickly generate video news reports. When urgent news events occur, reporters can input news scripts and related descriptions to rapidly generate accompanying video content, improving the efficiency and reach of news dissemination.
- Tourism Industry: Travel companies can combine 清影2.0 to generate detailed virtual tour videos. By inputting descriptions about tourist attractions, 清影2.0 can generate virtual guided tour videos, allowing tourists to experience tourist attractions panoramically online.
- E-commerce: E-commerce sellers can use 清影2.0 to generate product display videos. By inputting detailed descriptions and features of products, 清影2.0 can generate videos displaying the products, improving product appeal and sales.