Tongyi Wanxiang

Product Background and Launch Date

• Tongyi Wanxiang was officially launched on July 7, 2023, and is another masterpiece of AliCloud's continuous exploration and innovation in the field of AI technology.
• As a member of AliCloud's Tongyi series, Tongyi Wanxiang is designed for image creation, marking a solid step forward for AliCloud in the field of intelligent image generation.
• In January 2025, Tongyi Manphase was officially released as the latest version 2.1 of the Tongyi series, with significant improvements and optimizations mainly for the two main functions of video and image generation. This upgrade marks a further breakthrough in the field of multimodal generation, providing users with a richer and higher quality content creation tool.

Video Generationabilities

1. Efficient Codec Support::
   • Tongyi Wanxiang has introduced self-developed high-efficiency VAE (Variable Auto-Encoder) and DiT (Deep Image Transformation) architectures, which significantly enhance the spatio-temporal context modeling capability.
   • Supports efficient codecs for unlimited-length 1080P videos, greatly enhancing the flexibility and efficiency of video processing.
2. Chinese Text Video Generation::
   • For the first time, it realized the Chinese text video generation function and topped the VBench list, which demonstrated its leading position in video generation technology.
3. Artistic lettering and effects::
   • It supports one-click art character generation for Chinese and English videos, and provides a variety of video effects options, such as transition effects, particle effects and simulation effects, which greatly enrich the visual expression of the video.
4. Complex Transport Mirrors and Physical Simulation::
   • It supports complex mirroring, which can restore the real-world physical laws such as collision, rebound, cutting, extrusion, etc., making the video images more realistic and vivid.

Image Generation Capabilities

1. Vincennes Grouping Function::
   • Tongyi Wanxiang supports generating combinations of associated images through text descriptions, realizing the function of text-generated group images.
2. IC-LoRA image generation training method::
   • An IC-LoRA (Image Captioning with Low-Rank Adaptation) image generation training method is used, combined with the DiT architecture, to significantly enhance the text-to-image contextualization capability.
3. Characteristic stable continuum::
   • Through the splicing and joint description of multiple images, it can realize the combination generation between associated images and keep the features stable and continuous, which provides more creative space for designers and photographers.

Application Scenarios and Value

1. advertising production::
   • Tongyi Manphase's efficient video generation and rich special effect options make advertisement production more convenient and efficient.
2. animation::
   • Supports the function of complex lens transportation and physics simulation, which provides more realistic picture effect for animation production.
3. Creative Content Generation::
   • Whether you are a designer, photographer, or general user, you can easily personalize your art and design efficiently with Tongyi Wanxiang.

Technology Outlook

The upgrade of Tongyi Wanxiang marks the first step in Aliyun'sAI imageSignificant progress has been made in the field of video generation. In the future, with the continuous development of technology and the expansion of application scenarios, Tongyi Wanxiang is expected to play an important role in more fields and provide users with a richer and higher quality content creation experience.