AI enters Hengdian! Volcano Engine won a large order of 230 million yuan

Zhishi Author | Cheng Qian Editor | Li Shuiqing Zhixi News on October 14th, yesterday, the Dongyang Branch of the Jinhua Public Resource Trading Center in Zhejiang issued the "Announcement on the Winning Bid (Transaction) Result of the Hengdian Film and Television AI Service Platform Construction Project". The Dongyang Deep Space Engine and Volcano Engine consortium won the bid with a total of 230 million yuan. Among them, the quotation for computing power rental resources is 27 million yuan, and the remaining quotations are 203 million yuan. According to data from the enterprise information service platform Qichacha, Dongyang Deep Space Engine is indirectly wholly-owned by Douyin Co., Ltd. The project issued a bidding announcement on September 14 this year, with a total budget of 235 million yuan. Starting from the date of signing of the contract between the winning bidder and the purchaser, the project will be delivered and accepted in stages over three years. The specific project is to build a "data-model-application" trinity AI film and television technology system with the "Hengdian Film and Television Large Model" as the core. The core construction content includes Hengdian Film and Television large model service system, Hengdian Film and Television asset operation system, Hengdian Film and Television creative empowerment system, Hengdian Film and Television industry ecological management system, large model security, and computing resource leasing. The Hengdian film and television large model service system should be based on the leading general large model, and build a film and television exclusive large model through in-depth enhanced fine-tuning of vertical scenes and special scene adaptation. , focusing on breakthroughs in core capabilities such as lens language understanding, character emotion modeling, and multi-modal collaborative generation, covering script creation, storyboard generation, digital characters, sound generation, costume design, scene selection, special effects post-production and other core scenes, focusing on the creation of 7 vertical scene depth enhancement models covering the entire process of film and television production and 20 special scene enhancement models. In terms of data sets, the project requires building a high-quality data set system around the core scenes of the Hengdian film and television industry. The data set covers nine core directions, including real scenes of film and television bases, digital group performances, original IP digital artists, clothing supply chain, costumes/hairstyles/make-up of various dynasties, farming equipment and scene evaluations of past dynasties, forming a multi-modal, multi-scene, and high-standard data asset in the film and television domain. The project requires the construction of no less than 9 high-quality industry data sets, and also requires pre-training data sets.