a complete picture of everything

What is Tongyi Mansang?

Tongyi Wanxiang is an AI creative painting platform under AliCloud Tongyi, based on deep learning and Generative Adversarial Network (GAN) technology, focusing on Text-to-Image and Text-to-Video functions. It can generate images of various artistic styles according to the text descriptions input by users, support text-to-image, picture-to-picture, graffiti painting, virtual modeling, personal portraits and other picture creation scenarios, generate high-quality video content, and specialize in the simulation of complex sports scenes.

Main Functions of Tongyi Mansang

• Image Generation
  • Text to Image: The user inputs a text description, and the system generates a corresponding image based on the description, supporting a variety of styles to choose from.
  • Similar Image Generation: Based on existing images uploaded by users, new images are generated that are similar but with variations, providing users with more creative inspiration.
  • Image Style Migration: The user uploads an image and specifies a new art style, and the system converts the original image into a piece of art in the specified style.
  • Virtual Modeling and Personal Portraits: Generate realistic virtual model images or personal portrait photos by simple text or image input, suitable for fashion design, advertising and promotion, and other fields.
• Video Generation
  • Text Generation Video: Enter a text description to generate a video that matches the description.
  • Image Reference Video Generation: Generate videos based on images that are related to the style or content of the image.
  • Video redrawing: Style conversion or content adjustment of existing videos.
  • Video partial editing: Edit localized areas of the video, such as replacing, adding or removing elements.
  • Video Background and Duration Extension: Extend the background range of a video or extend the duration of a video.
• Multi-modal inputs: Supports text, images, video frames, masks, and a variety of control signals (e.g., human gesture, motion optical flow, structure hold, depth map, layout, line drawings, etc.) as inputs.
• Intelligent Expansion: Before generating an image or video, you can use the Smart Amplify feature to add details to the inspired description.
• Multiple model options: Provide different versions of models, such as Wanxiang 2.1 Professional Edition, Wanxiang 2.1 Extreme Edition, etc. Users can choose the suitable model according to their needs.

Tongyi Mansang official website address

Official website address: https://tongyi.aliyun.com/wanxiang/

How to use Tongyi Universal Phase

• Register for an account: Visit the official website of Tongyi Mansang and register an account using your cell phone number.
• Selection FunctionsAfter logging in, you will enter the page of Tongyi Wanxiang, and you can select the functions of "Text Image Generation", "Similar Image Generation", "Image Style Migration", and "Applications Plaza" and other functions.
• Text to ImageIn the "Text to Image" interface, enter a text description, such as "landscape painting", click "Generate Creative Paintings", and the system will generate the corresponding image.
• If you are not satisfied, you can click "Generate Again" or "Reuse Ideas" to add elements. After generating a satisfactory drawing, you can like it, generate a similar picture, download the original picture or favorite it.
• Similar Image GenerationClick on "Generate Similar Image", select a picture or enter text, click "Generate Similar Painting", and the system will generate an image similar to the original one.
• Image Style MigrationClick "Image Style Migration", select the original image and the style image, click "Generate", the system will convert the original image to the specified style.
• Application PlazaIn the application square, select the virtual model function, upload the product display picture, select the reserved area, configure the model image and scene, and click "Generate" to get the model blockbuster.
• Video GenerationLogin to the official website of Tongyi Wanxiang, find the "Video Generation" portal, and click "Text to Video" or "Picture to Video". Input text prompts or upload pictures, adjust parameters such as video ratio, etc., click "Generate" and wait for about 2-5 minutes.

Product Advantages of Tongyi Vanphase

• Multimodal information input: Tongyi Manphase supports text, image, video, mask, and a variety of control signals (such as depth map, optical flow, layout, grayscale, line art, and gesture) as inputs, improving the controllability of video generation.
• A single model for harmonization: It is a unified model for video editing that enables a variety of tasks such as image referencing, video redrawing, and local editing.
• Efficient Model Architecture: Based on advanced deep learning and Generative Adversarial Networks (GANs), Tongyi Wanxiang uses multi-stage progressive training combined with Flow Matching framework to achieve efficient and stable convergence of 14B parameter models.
• Spatio-temporal Variational Auto-Encoder (Wan-VAE): Designed for video generation, it improves temporal compression, reduces memory usage, and ensures temporal causality.
• Video Diffusion DiT: Adopting the mainstream video DiT structure, we effectively model the long-duration spatio-temporal dependency through Full Attention mechanism to generate spatio-temporally consistent high-quality videos.
• High-quality content generation: It generates realistic visual effects that strictly follow the rules of physics to enhance the pixel quality and realism of the picture.
• Complex Motion Processing: Specializes in handling large-scale body movements and complex rotations, such as figure skating and swimming scenes, maintaining body coordination and realistic movement trajectories.

Application Scenarios of Tongyi Manxiang

The application scenarios of Tongyi Vanguard include: creative inspiration, when artists encounter inspiration bottlenecks in the process of creation, they can input keywords or descriptions to let Tongyi Vanguard generate relevant images; promotional posters and advertisement copy, to quickly generate promotional posters, prototype drawings of products, or advertisement copy; instructional video production, teachers can generate vivid animation videos and experimental demonstration videos according to the content of the teaching; conceptual maps and scene design, to generate conceptual maps and scene designs for film and television, to help film and television production teams quickly conceive ideas. Concept map and scene design, generating movie and TV concept maps, scene design, etc., helping movie and TV production teams to quickly conceptualize. Graphics and Script Creation: Provide inspiration support for short videos, article graphics or scripts to meet the needs of social media content creation.