By the end of this year, the technological competition among major domestic video platforms is extremely fierce. The first Chinese character video technology launched by Tongyi Wanxiang has aroused enthusiastic response in the field of AI creation. Immediately afterwards, Conch AI launched a new subject reference function, and this innovation has brought AI video creation to a new level. Users only need to upload a picture to allow any character to move freely in various scenes, which not only maintains the character's high restoration, but also gives creators great freedom.
On social media, creators show their works, from the scene of Spider-Man riding a motorcycle to the interaction between the Dragon Mother and the Forest Wolf in Game of Thrones, these characters not only maintain a high degree of restoration, but also show the Excellent stability.
Compared with traditional graphic video technology, Conch AI's "subject reference" function adopts a new technical path. This technology not only allows static images to produce local actions, but can generate complete video clips based on text prompt words (prompt), greatly enriching the expressiveness of the video content.
From the official demonstration video, we can see that after the user uploads a character photo, the system can not only accurately ensure the facial features of the person holding the character, but also allow him to perform various natural and coherent actions in the scene, such as skateboarding and riding a bicycle. , raising eyebrows or frowning, etc., the details of facial expressions are also handled very well.
In terms of technical implementation, Conch AI chose a route based on image reference rather than traditional LoRA technology. This strategy allows users to accurately identify and restore character features based on just one picture without preparing a large amount of material. In addition, the generation speed has also been significantly improved, shortening from the previous long wait to the second-level response.
The commercial potential of this technology is equally huge. According to Statista, the market size of generative AI products in the advertising marketing field is expected to reach US$107.5 billion by 2028. The "Psychological Reference" function just meets this market demand, especially in product advertising production, where you only need one model picture to generate multiple versions of video content for different product scenarios.
As a startup that has been established for only 3 years, MiniMax has established a leading position in the global AI video generation through Conch AI. At present, only Conch AI, Keling AI and Jimeng AI have shown real technical strength and development potential around the world, which marks the rise of Chinese companies in the global AI video generation field.
Looking ahead, Conch AI plans to expand the "subject reference" function to more areas such as multiple people, objects and scenes, further unleash creative potential. As AI video generation is about to enter the second half, the emergence of this technology undoubtedly brings more possibilities to creators, making the vision of "every idea is a blockbuster" gradually become a reality.
Experience address: https://hailuoai.com/video/create