Meitu Releases AI Vision Model 4.0, AI Video Generation Set for Major Breakthrough
-
On December 5-6, the Meitu Creativity Conference themed 'Future AI Design' was held in Xiamen. Meitu Company released its self-developed AI vision model MiracleVision (Fantasy Intelligence) version 4.0, focusing on AI design and AI video.
In terms of AI design, four new capabilities have been added: vector graphics, text effects, intelligent layering, and intelligent typesetting. A new visual model store has also been launched, supporting the creation of personal visual models to help designers achieve unique artistic pursuits.
In terms of AI video, four new capabilities have been introduced: text-to-video, image-to-video, video camera movement, and video-to-video. According to reports, MiracleVision's AI video capabilities can already be integrated into industry workflows, especially in e-commerce and advertising.
To animate images while showcasing creativity, Meitu's large model team has integrated precisely calculated motion modules into its image generation capabilities. The system is trained on high-quality video datasets, ensuring that version 4.0 accurately understands semantics, generates smooth video content, and supports up to 4K ultra-high-definition resolution.
MiracleVision has officially opened its commercial API to the public. MiracleVision 4.0 will be gradually rolled out to Meitu's products starting January 2024. Users will be able to experience it in apps like Meitu Xiuxiu, Meiyan Camera, Wink, Meitu Design Studio, and WHEE.
Wu Xinhong, founder, chairman, and CEO of Meitu, stated at the conference that MiracleVision has already served 1.745 million e-commerce users, generating a cumulative total of 104 million AI product images. Currently, Meitu has partnered with dozens of e-commerce platforms and brand clients, including Weidian, Shihuo, Deli, and Feiyu.
Additionally, MiracleVision will collaborate with the large model developed by Weitu AI, a company led by Meitu's angel round investment. Wu Xinhong mentioned that once MiracleVision's cognitive capabilities improve, Meitu users will be able to use natural language more effortlessly during collaborative AI creation, enhancing both efficiency and experience.
AI design was a key focus at the conference. Wu Xinhong announced that Meitu has formed a deep strategic partnership with Zcool (a popular Chinese designer interaction platform) to jointly launch the 'Emerging AI Designer' program. This initiative will provide designers with AI design tools, learning platforms, professional communities, and business opportunities. "We will offer better services to help designers showcase their innovative talents in the AI era," said Wu Xinhong.
Subsequently, Meitu and Zcool jointly released the '2023 AI Design Practice Report'. Senior Vice President Chen Jianyi provided a brief introduction to the report on-site. Key findings indicate that the adoption rate of AI design tools among individuals in China has exceeded expectations, though 30% of respondents still abandon usage due to issues like cost, complexity, or unsatisfactory results. Enterprise adoption remains relatively low, with only 8.4% of companies currently utilizing AI tools.
The research report found that the internet industry exhibits the highest enthusiasm for AI design tools, followed by sectors such as food & beverage, footwear & apparel, and beauty & cosmetics that utilize AI tools to drive commercial transformation.
Overseas AI tools currently have significantly higher adoption rates than domestic alternatives. However, the report suggests Chinese AI tools possess substantial potential, with key advantages including lower accessibility barriers, native Chinese language support, and alignment with Asian aesthetics.
The three major pain points of current AI design tools are identified as: imprecise control, generation flaws, and copyright disputes. Regarding workflows, AI primarily serves as a creative assistant at this stage, still requiring integration with traditional tools to deliver finalized designs.
The report makes five major trend predictions:
- AI will lead to the generalization of design, making creativity and aesthetics the core competitiveness of designers.
- AI design capabilities and coverage categories will become more comprehensive, becoming a standard in design workflows.
- AI video generation will experience a major boom in 2024.
- AI will shine in the marketing field, with AI-driven marketing cases that go viral nationwide just around the corner.
- The usage rate of domestic AI tools in the Chinese market is expected to surpass that of overseas tools by the end of 2024.