news

on september 25, tang jiayu, co-founder and ceo of shengshu technology, announced at the baidu cloud intelligence conference that vidu, a video big model under shengshu technology, officially opened its api (application programming interface) and was simultaneously connected to the baidu smart cloud qianfan big model platform, becoming the first video big model to be connected to the platform.

as one of the earliest teams in china to develop multimodal general large models, shengshu technology jointly released the video large model vidu with tsinghua university in april this year.

in june, shengshu technology completed a pre-a round of financing worth hundreds of millions of yuan, led by baidu and beijing artificial intelligence industry investment fund, and followed by zhongguancun science park company and qiming venture partners. at that time, shengshu technology said that it would continue to train and improve model capabilities based on baidu's baige ai heterogeneous computing platform, and gradually open model services through baidu intelligent cloud qianfan platform.

according to tang jiayu, the architecture adopted by vidu is the u-vit architecture developed purely by the team. it is the world's earliest diffusion transformer fusion architecture, earlier than sora's dit architecture. this has laid an important foundation for general generation tasks.

tang jiayu, co-founder and ceo of shengshu technology. image source: provided by the company

he also said that the vidu model has the ability to generate videos from text and images, and supports chinese and english command input; in terms of duration, vidu can support the generation of up to 32s video with one click at the model level; in terms of picture quality, vidu can output up to 1080p resolution.

shengshu technology said that currently, enterprises and institutions in the film, television, animation, advertising and other industries generally have large-scale video output needs. the opening of vidu api will help these companies reduce costs and increase efficiency in the video production process and stimulate creativity. at the same time, for many developers, the opening of vidu api also provides an important foundation for exploring ai 2.0 applications.

shengshu technology believes that the highly personalized and automated content creation capabilities of large video models will give companies new competitiveness in marketing, brand promotion, content innovation and other scenarios. the introduction of video models will become a key factor in improving the competitiveness of the creative industry. based on this background, opening the vidu api has become an important strategic measure for shengshu technology to further promote its commercialization layout.

since 2024, the competition for video models has become increasingly fierce. on september 24, bytedance just announced the release of two large models, doubao video generation-pixeldance and doubao video generation-seaweed. prior to this, there was already kuaishou's video generation large model "keling" in the industry; at the same time, alibaba's tongyi wanxiang announced a comprehensive upgrade in september and released a new video generation model; meitu xiuxiu, an old player in the image track, also announced in september that its meitu qixiang large model (miraclevision) video generation capabilities were fully upgraded.

daily economic news